Method, device, and recording medium for image encoding / decoding

Abstract

A method, a device, and a recording medium for image encoding / decoding in which prediction is performed are disclosed. The encoding / decoding method comprises the steps of: deriving a prediction mode for a target block from among various available prediction modes; configuring a reference sample for prediction of the target block by using selection, filtering, sampling, interpolation, and the like for the reference sample; and performing prediction on the target block by using the reference sample and the derived prediction mode. The prediction mode may include an intra prediction mode to which an extrapolation filter is applied.

Description

Method, apparatus, and recording medium for video encoding / decoding

[0001] The present invention relates to a method, apparatus, and recording medium for image encoding / decoding. Specifically, the present invention discloses a method, apparatus, and recording medium for image encoding / decoding using intra-prediction.

[0002] The present invention claims the benefit of the filing dates of Korean Patent Application No. 10-2025-0032348 filed on March 12, 2025 and Korean Patent Application No. 10-2026-0044912 filed on March 12, 2026, the entire contents of which are incorporated herein.

[0003] With the continuous development of the information and communication industry, services providing video through broadcasting and the Internet have spread globally.

[0004] Users demand videos with higher resolution and quality. To meet these user demands, video encoding and decoding technologies suitable for such videos are required. Video encoding technology can generate compressed video by compressing the video representing the images to have a smaller amount of data. Video decoding technology can generate reconstructed images using the compressed video.

[0005] Regarding video encoding and decoding technologies, various techniques exist, such as segmentation, prediction, transformation, quantization, filtering, and entropy encoding and decoding. By introducing, modifying, improving, and combining these various techniques, video and images can be compressed, transmitted, and stored more effectively.

[0006] One embodiment may provide an apparatus, method, and recording medium for image encoding / decoding using intra-prediction.

[0007] One embodiment may provide an apparatus, method, and recording medium for image encoding / decoding using an extrapolation filter.

[0008] One embodiment may provide an apparatus, method, and recording medium for video encoding / decoding intra-template matching prediction.

[0009] A video decoding method is provided, comprising: a step of inducing a prediction mode for a current block in one aspect; a step of configuring a reference sample for the prediction of the current block; and a step of performing the prediction for the current block using the reference sample and the induced prediction mode.

[0010] The above prediction can be performed using an intra-prediction mode in which an extrapolation filter is applied to the current block.

[0011] In an intra-prediction mode where the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block can be applied to the chroma block.

[0012] Correction can be performed on the prediction block to which the extrapolation filter applied to the above current block is applied.

[0013] The extrapolation filter for the current block above can be inherited from the extrapolation filter of the restored block.

[0014] The extrapolation filter of the current block can be derived using samples within the restored region surrounding the current block.

[0015] Correction can be performed on the coefficient values ​​of the extrapolation filter of the current block.

[0016] In another aspect, an image encoding method is provided, comprising: a step of deriving a prediction mode for a current block; a step of configuring a reference sample for the prediction of the current block; and a step of performing the prediction for the current block using the reference sample and the deriving prediction mode.

[0017] The above prediction can be performed using an intra-prediction mode in which an extrapolation filter is applied to the current block.

[0018] In an intra-prediction mode where the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block can be applied to the chroma block.

[0019] Correction can be performed on the prediction block to which the extrapolation filter applied to the above current block is applied.

[0020] The extrapolation filter for the current block above can be inherited from the extrapolation filter of the restored block.

[0021] The extrapolation filter of the current block can be derived using samples within the restored region surrounding the current block.

[0022] Correction can be performed on the coefficient values ​​of the extrapolation filter of the current block.

[0023] In another aspect, a computer-readable recording medium is provided for storing a bitstream generated by a video encoding method.

[0024] In another aspect, a computer-readable recording medium for storing a bitstream is provided, wherein the bitstream comprises prediction information, a prediction mode for a current block is derived using the prediction information, a reference sample for the prediction for the current block is configured, and the prediction for the current block is performed using the reference sample and the derived prediction mode.

[0025] The above prediction can be performed using an intra-prediction mode in which an extrapolation filter is applied to the current block.

[0026] In an intra-prediction mode where the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block can be applied to the chroma block.

[0027] Correction can be performed on the prediction block to which the extrapolation filter applied to the above current block is applied.

[0028] The extrapolation filter for the current block above can be inherited from the extrapolation filter of the restored block.

[0029] An apparatus, method, and recording medium for video encoding / decoding using intra-prediction are provided.

[0030] An apparatus, method, and recording medium for image encoding / decoding using an extrapolation filter are provided.

[0031] An apparatus, method, and recording medium for video encoding / decoding of intra-template matching prediction are provided.

[0032] FIG. 1 shows a system for video coding according to one embodiment.

[0033] Figure 2 shows a segmentation structure of an image according to one embodiment.

[0034] Figure 3 shows the structure of an intra prediction according to one embodiment.

[0035] FIG. 4 shows the structure of an inter prediction to explain an inter prediction process according to one embodiment.

[0036] FIG. 5 shows the order of addition of spatial candidates to the candidate list according to one embodiment.

[0037] Figure 6 shows a plurality of in-loop filters according to one example.

[0038] Figure 7 shows the structure of entropy encoding and entropy decoding according to one example.

[0039] FIG. 8 is a flowchart of a method for predicting a target block and a method for generating a bitstream according to one embodiment.

[0040] FIG. 9 is a flowchart of a method for predicting a target block using a bitstream according to one embodiment.

[0041] Figure 10 is a flowchart of a prediction method according to one example.

[0042] Figure 11 is a flowchart of an encoding method using an extrapolation filter according to one example.

[0043] Figure 12 is a flowchart of a decoding method using an extrapolation filter according to one example.

[0044] FIGS. 13a to 13c show the forms of extrapolation filters according to one example.

[0045] FIG. 14 shows specific locations within a luminance block and a color difference block that is the subject of prediction according to one example.

[0046] FIGS. 15a to 15y show the positions of the size and coefficients of an extrapolation filter according to one example.

[0047] FIGS. 16a to 16c show the restoration regions for an extrapolation filter according to one example.

[0048] Figure 17 shows a reference area for the current block and extrapolation filter according to one example.

[0049] Figure 18 illustrates the concept of an intra-block copy mode according to one example.

[0050] Figure 19 illustrates the concept of template matching prediction according to one example.

[0051] Various modifications may be applied to the present invention. Additionally, the present invention may have various embodiments. Specific embodiments are described by the drawings and the detailed description.

[0052] Specific embodiments are not intended to limit the invention to specific embodiments, and it should be understood that all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention are included as embodiments of the invention.

[0053] The embodiments are described in sufficient detail to enable those skilled in the art to practice the embodiments. It should be understood that the various embodiments are different but need not be mutually exclusive. For example, it should be understood that the shapes, structures, and characteristics described in relation to one embodiment may be applied to or implemented in other embodiments without departing from the spirit and scope of the invention. It should also be understood that the location or arrangement of components within one embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the exemplary embodiments is limited only by the appended claims and all equivalents to the scope claimed by such claims, provided that they are appropriately described.

[0054] The detailed description of the embodiments described below may refer to the drawings relating to the embodiments. Descriptions described in the drawings or descriptions represented by the drawings may be considered part of the detailed description. In the drawings, similar reference numerals may refer to the same or similar functions for various aspects. Dependencies between components may not be limited to those depicted in the drawings.

[0055] In the embodiments, singular expressions may include plural expressions and may be limited to and / or limited to plural expressions unless the context clearly excludes plural expressions. That is to say, in the embodiments, expressions such as 'at least one' and 'one or more' may be replaced with 'plural'. Terms such as ' / ', 'and / or', 'at least one of' and 'one or more of' described for plural items may mean 1) one of the plural items, 2) some of the plural items, 3) a combination of some of the plural items, or 4) a combination of the plural items. Additionally, plural expressions may be replaced with singular expressions. Plural may mean an integer of 1, 2, 3, 4, or 5 or more.

[0056] In the embodiments, numbered terms such as 'first' and 'second' may be used to describe various components. These terms are used solely for the purpose of distinguishing one component from another and do not limit the components. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component.

[0057] The statement that a first component transmits (or provides) information to a second component may mean that the first component directly transmits information to the second component, or it may mean that the first component transmits information to the second component through another third component. Here, the information received (or acquired) by the second component may be information transmitted by the first component, or information generated by applying a specific processing to information transmitted by the first component.

[0058] The components of the embodiments may be illustrated independently to represent different characteristic functions, and this does not imply that each component corresponds to a separate hardware or a single software unit. That is, the components of the embodiments may be classified and enumerated for convenience of description. Two or more components described in the embodiments may be regarded as a single component. Furthermore, a single component described in the embodiments may be separated into multiple components that perform the functions of the said component separately. Embodiments in which such components are integrated and embodiments in which components are separated are also included within the scope of the present invention, provided that they do not depart from the essence of the invention.

[0059] The terms used in the embodiments are used merely to describe specific embodiments and are not intended to limit the invention. In the embodiments, terms such as "comprising" or "having" indicate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the embodiments. The existence or addition of other features, numbers, steps, actions, components, parts, or combinations thereof not explicitly described in the embodiments is not excluded by these terms. That is, the description of a specific component of an embodiment as "comprising" does not exclude components other than the specific component, and means that additional components may also be included within the scope of the embodiments or the technical concept of the invention.

[0060] Some of the components of the embodiments may be optional components that are not essential for performing the essential functions of the invention. Such optional components may be used to enhance performance. The embodiments may be implemented as a structure comprising only the essential components required to realize the essence of the embodiments, excluding the optional components. Such a structure is also included within the scope of the embodiments.

[0061] In the following, embodiments are described in detail with reference to the attached drawings so that a person skilled in the art can easily implement the embodiments. In describing the embodiments, if it is determined that a detailed description of related known configurations or known functions could obscure the gist of this specification, such detailed description is omitted. Additionally, the same reference numerals are used for identical components within the drawings, and redundant descriptions of identical components are omitted.

[0062]

[0063] Replacement of terms in the examples

[0064] Below, terms listed in a single line may be used with the same meaning in the examples and may be used interchangeably in the examples.

[0065] - 'one or more', 'at least one'

[0066] - 'two or more', 'a plurality of', 'multiple', 'multiple'. (In the examples, 'one or more' or 'at least one' may be further limited to 'two or more', 'multiple', or 'multiple'.)

[0067] - 'Information', 'Signal'

[0068] - 'value', 'predefined value', 'specific value', 'threshold', 'threshold value', 'baseline value', 'reference value'

[0069] - 'statistical value', 'statistics value'

[0070] - 'indicator', 'index', 'index', 'flag', 'information'

[0071] - 'encoder', 'encoding apparatus'

[0072] - 'decoder', 'decoding apparatus'

[0073] - 'Entropy encoding', 'encoding', 'encoding'

[0074] - 'Entropy decoding', 'decoding', 'decoding'

[0075] - 'Coding', 'Encoding and / or decoding'

[0076] - 'video', 'moving picture', 'image', 'picture', 'picture', 'frame', 'screen'

[0077] - 'Reference picture', 'Reference video'

[0078] - 'Reference Picture List (RPL)', 'Reference Video List'

[0079] - 'original', 'input', 'source'

[0080] - 'Block', 'Unit', 'Signal'

[0081] - 'square', 'square shape'

[0082] - 'pixel', 'pixel', 'sample', 'pel'

[0083] - 'region', 'area', 'part', 'segment'

[0084] - 'partition', 'split', 'divide'

[0085] - 'quad', 'quaternary'

[0086] - 'Luma component', 'Luma', 'luminance component', 'luminance', 'Y'

[0087] - 'Chroma component', 'Chroma', 'chrominance', 'chrominance component', 'Cb and Cr', 'Cb or Cr', 'Cb', 'Cr', 'U and V', 'U or V', 'U', 'V'

[0088] - 'target', 'current' (e.g., target block and current block, or target image and current image)

[0089] - 'neighbor', 'neighboring', 'adjacent', 'neighbor / neighboring' (e.g., neighbor block, adjacent block, and neighboring block)

[0090] - 'collocated', 'COL'

[0091] - 'reconstruction', 'reconstruction', 'decoding'

[0092] - 'reconstructed', 'reconstructed', 'decoded'

[0093] - 'Difference', 'Difference', 'Difference', 'Error', 'Residual', 'Residual'

[0094] - 'Largest Coding Unit (LCU)', 'Coding Tree Unit (CTU)'

[0095] - 'inter', 'inter-screen'

[0096] - 'Inter prediction', 'inter prediction', 'motion compensation'

[0097] - 'Inter Mode', 'Inter Prediction Mode', 'Inter-frame Mode', 'Inter-frame Prediction Mode'

[0098] - 'Bi-prediction', 'bi-directional prediction', 'inter-bi-prediction', 'bi-directional inter-prediction'

[0099] - 'Motion Information', 'Motion Vector', 'Block Vector', 'Predicted Motion Vector', 'Advanced Motion Vector Prediction (AMVP)'

[0100] - 'List', 'Candidate List'

[0101] - 'spatial candidate', 'spatial merge candidate'

[0102] - 'temporal candidate', 'temporal merge candidate'

[0103] - 'prediction motion vector candidate', 'motion vector predictor'

[0104] - 'Prediction method', 'Prediction mode'

[0105] - 'Intra', 'Inside the screen'

[0106] - 'Intra prediction', 'Intra prediction'

[0107] - 'Intra mode', 'Intra prediction mode', '(Intra) predictor'

[0108] - 'Dequantization', 'Scaling'

[0109] - 'Quantization matrix', 'Scaling list'

[0110] - 'Quantization matrix coefficients', 'Matrix coefficients'

[0111] - 'transform coefficient level', 'quantized level', 'quantized coefficient', 'quantized transform coefficient', 'quantized transform coefficient level'

[0112] - 'dequantized coefficient', 'dequantized transform coefficient'

[0113] - 'Scanning type', 'Scanning direction'

[0114] - 'directional mode', 'angle mode', 'angular mode', 'intra-prediction mode'

[0115] - 'Intra-prediction mode (mode) number', 'Intra-prediction mode (mode) index', 'Intra-prediction mode (mode) value', 'Intra-prediction mode (mode) angle', 'Intra-prediction mode (mode) direction', 'Intra-prediction direction (mode) number', 'Intra-prediction direction (mode) index', 'Intra-prediction direction (mode) value', 'Intra-prediction direction (mode) angle'

[0116] - 'Merge Mode', 'Motion Merge Mode'

[0117] - 'Geometric Partitioning Mode (GPM)', 'Triangle Partitioning Mode'

[0118] - 'Refinement', 'Correction'

[0119] In addition to the terms exemplified above, terms having the same meaning according to the ordinary knowledge of the technical field may be used interchangeably in the embodiments.

[0120]

[0121] Information and range of values ​​of information described in the embodiments

[0122] In the embodiments, information may include a constant, a flag, an index, a variable, a coding parameter, an element, a syntax element, motion information, an attribute, an entity, an object, and data, etc. That is to say, the term 'information' may be interchangeable with 'data', 'flag', 'index', 'variable', 'element', 'syntax element', 'motion information', 'attribute', or 'entity'.

[0123] Information can have one of multiple values. 'The nth value' can refer to the nth value among multiple values.

[0124] For example, the first value can represent '0' or (logical) false. The second value can represent '1' or (logical) true. Or, the first value can represent '1' or (logical) true. The second value can represent '0' or (logical) false. The second value may mean a value other than the first value.

[0125] In the embodiments, "when specific information is false" may mean "when specific information is not true." A specific embodiment described as being performed when specific information is true may not be performed when specific information is false. "When specific information is true" may mean "when specific information is not false." An embodiment described as being performed when specific information is false may not be performed when specific information is false.

[0126] A flag may be information having a value of either '0' or '1'. In the embodiments, the values ​​'0' and '1' of the flag may be replaced with '1' and '0', respectively. For example, information indicating whether a specific process is performed or information indicating whether a specific process is applied may be considered as a flag.

[0127] When a variable such as i or j is used to represent a row, column, or index, the variable may be an integer between 0 and n - 1 inclusive. Or, the variable may be an integer between 1 and n inclusive. Here, n may be the number of rows, the number of columns, or the number of entities pointed to by the index.

[0128]

[0129] Concepts related to coding

[0130] Concepts related to coding are explained below. The descriptions disclosed below may be applied to embodiments.

[0131] Predefined value: The predefined value may refer to a value commonly used by the encoding device and the decoder. For example, the predefined value may be interpreted as being limited to a fixed value. Alternatively, the predefined value may be a value shared by the encoding device and the decoder through signaling. Alternatively, the predefined value may be a value derived through the same procedure in the encoding device and the decoder so that the encoding device and the decoder have a common value. Alternatively, the predefined value may be a common value possessed by the encoding device and the decoder.

[0132] In the encoding device and the decoding device, the 'value derived through the same procedure' may include the value derived through the same procedure and / or the same conditional statement for the same value and / or the same information in the encoding device and the decoding device.

[0133] The description of the predefined values ​​above may also apply to predefined information. In the descriptions above, 'value' may be replaced with 'information'.

[0134] Availability: The availability of specific modes for a specific target may mean that a selected mode among the specific modes is used for that specific target. Other modes belonging to the category of specific modes may be non-available modes. Non-available modes may not be used for a specific target. The above description of specific modes may also apply to other specific information. In the above descriptions, 'mode' may be replaced with 'information'.

[0135] Adjacency: 'Direction' for 'First Object'. 'Second Object' may refer to a 'Second Object' adjacent to the 'Direction' corner / face of the First Object. For example, the 'Top-left Block' for a 'Target Block' may be a block adjacent to the top-left of the Target Block. Here, the 'First Object' may be a Target Unit, Target Block, or Target Sample. 'Direction' may be one of left-above, above, right-above, left, right, left-below, below, and right-below. The 'Second Object' may be a Unit, Block, or Sample. For the directions of top-left, top-right, bottom-left, and bottom-right, the corner of the First Object and the corner of the Second Object may be diagonally adjacent. For the directions of top, left, right, and bottom, one face of the First Object and one face of the Second Object may be in contact with each other.

[0136] - For example, the block adjacent to the top-left of the target block may be the block adjacent to the top of the block adjacent to the left of the target block. The block adjacent to the top-right of the target block may be the block adjacent to the right of the block adjacent to the top of the target block. The block adjacent to the bottom-left of the target block may be the block adjacent to the bottom of the block adjacent to the left of the target block.

[0137] Coding: Coding can refer to encoding and / or decoding of an image.

[0138] Signal: A signal can represent information about an image, unit, or block. A specific signal can represent a specific image, a specific unit, or a specific block.

[0139] Image: An image can refer to a single picture constituting a video, or it can represent the video itself. For example, "encoding and / or decoding of an image" can mean "encoding and / or decoding of a video," or it can mean "encoding and / or decoding of one of the images constituting a video."

[0140] - An image can refer to the entirety of a picture, or it can refer to a part of a picture, such as a block.

[0141] Target image: The target image may be an encoding target image that is the subject of encoding and / or a decoding target image that is the subject of decoding. Additionally, the target image may be an input image processed by an encoding device and a restored image processed by a decoding device. The target image may be an image containing a target block.

[0142] Subpicture: A picture can be divided into one or more subpictures.

[0143] - A subpicture may be a square or rectangular area within the picture. A subpicture may include one or more CTUs.

[0144] - A subpicture may include one or more slices and / or one or more tiles. For example, a subpicture may consist of one or more slice rows and one or more slice columns. Alternatively, each subpicture may consist of one or more tile rows and one or more tile columns.

[0145] - A subpicture may include one or more slices that collectively cover a rectangular area within the picture. Accordingly, the boundary of each subpicture can always be the boundary of a slice. Additionally, each vertical subpicture boundary can always be the boundary of a vertical tile.

[0146] Slice: A slice may include one or more tiles within a picture. A slice may consist of one or more rows of tiles and one or more columns of tiles.

[0147] Tile: A tile can be a square or rectangular area within a picture. A tile can contain one or more CTUs. A picture can be divided into one or more tile rows and one or more tile columns.

[0148] CTU: An image can be divided into multiple Coding Tree Units (CTUs).

[0149] - A CTU may include one Y Coding Tree Block (CTB) and at least one of a Cb CTB and a Cr CTB associated with the Y CTB, and may include information for each CTB. The information may include syntax elements.

[0150] - Each CTU may be partitioned using one or more partitioning methods to form sub-units such as Coding Units (CU), Prediction Units (PU), and Transform Units (TU). One or more partitioning methods may include Quad Tree (QT) partitioning, Binary Tree (BT) partitioning, and Ternary Tree (TT) partitioning. Additionally, each CTU may be partitioned using Multi-Type Tree (MTT) partitioning, which uses a combination of multiple partitioning methods.

[0151] CTB: CTB can refer to one of Y CTB, Cb CTB, and Cr CTB.

[0152] Unit: A unit can be determined for specific processing in coding. A unit may be information about a specific region within an image. For specific processing in coding, the image may be recursively divided into multiple parts. A unit may represent the region to which the specific processing is applied and information about the aforementioned region.

[0153] - The unit type may represent a specific process applied to the unit. Depending on the unit type, a specific process may be applied to the unit. The 'specific' unit may be a unit for the process named 'specific' in the coding. For example, the unit may be at least one of the source unit, CTU, coding unit, prediction unit, residual unit, restored residual unit, transformation unit, and restored unit.

[0154] - A unit may include a parameter set, a header, a brick, a CTU, a CU, a PU, and a TU. A unit may include a VPS, an SPS, a PPS, an APS, and a DPS. A unit may include a picture header, a subpicture header, a slice header, a tile group header, and a tile header.

[0155] - A unit may include samples having a two-dimensional form or arrangement. In this respect, a 'unit' may mean a 'block'. For example, a block may be at least one of an original block, a CTB, a coding block (CB), a prediction block (PB), a residual block, a restored residual block, a transform block (TB), and a restored block. For example, a partition of a unit may mean a partition of a block corresponding to the unit.

[0156] - In the embodiments, the description applied to one of the various units described above may also be applied to other units.

[0157] - A unit may include syntactic elements. That is to say, a block and a syntactic element for the block may be combined and referred to as a 'unit'. In the embodiments, the description that a specific unit is a specific method / mode may mean that a specific method / mode is performed in a specific unit, and may mean that a specific unit has a syntactic element representing a specific method / mode.

[0158] - A block may be an MxN array of samples. Here, M and N may represent positive integer values, and a block often refers to a two-dimensional array of samples. The current block may refer to the encoding target block that is the subject of encoding during encoding, or the decoding target block that is the subject of decoding during decoding. Additionally, the current block may be at least one of a coding block, a prediction block, a residual block, a transformation block, or a restoration block. Blocks may have various sizes and shapes. For example, the shape of a block may be one or more of a tetragon, a rectangular, a square, a rectangle where the width differs from the height (i.e., an oblong), a trapezoid, a triangle, a right-angled triangle, and a pentagon. Here, the width and height of the rectangle may differ from each other. Additionally, the shape of a block may include other geometric figures that can be represented in two dimensions. For example, the shape of the block may be a square or a pentagon defined by subtracting the area of ​​a right triangle from the area of ​​a rectangle. Here, the right-angled vertex of the right triangle may be one of the vertices of the rectangle. Additionally, the shape of the block may be a combination of two or more of the aforementioned shapes. Additionally, the shape of the block may be the remainder of one of the aforementioned shapes after another shape has been subtracted.

[0159] - In the embodiments, the rectangle may be limited to a non-square rectangle. When the shape of a specific object in the embodiments is described as a rectangle, this description may additionally imply that the width and height of the specific object are different from each other.

[0160] - In the embodiments, the block may be limited to at least one of a vertically oriented block and a horizontally oriented block. A vertically oriented block may mean a block in which the vertical length is greater than the horizontal length. A horizontally oriented block may mean a block in which the horizontal length is greater than the vertical length.

[0161] - The unit may include a luma component block (i.e., a Y block) and two chroma component blocks (i.e., at least one of a Cb block and a Cr block), and may include information for each block. The information may include syntax elements.

[0162] - The unit information may include the unit type, unit size, unit depth, unit encoding order, and unit decoding order.

[0163] Target Unit: The target unit may be a block that is the target of encoding, an encoding target unit, and / or a decoding target unit that is the target of decoding. The target unit may be a specific region within the target picture to which one or more specific processing steps of coding are applied. By applying a specific processing step to the target unit, a unit of a specific type may be generated. Alternatively, the target unit may represent a unit having a specific type for a specific processing step of coding.

[0164] Depth: A block can be hierarchically divided into multiple sub-blocks with depth according to a tree structure. The multiple sub-blocks created by the division of a block can be referred to as partitions.

[0165] - The block depth can represent the level of the node corresponding to the block when the blocks constituting the image are represented as a tree structure. Alternatively, the block depth can represent the number of divisions applied until the block is determined. The block depth can increase by 1 as the block is further divided.

[0166] - In a tree structure, the root node can be considered to have the smallest level, and the leaf node the largest level. The root node may be the top node of the tree structure and may correspond to the first undivided block. The level of the root node may be 0 or 1. When the level of the root node is 0, a node with level 1 may represent the block determined by the first block being divided once. A node with level n may represent the block determined by the first block being divided n times. A leaf node may be the lowest node of the tree structure. A leaf node may be a node that cannot be further divided. The depth of a leaf node may be a predefined maximum depth. For example, the maximum depth may be a positive integer such as 3. The root node may represent a CTU. A leaf node may represent at least one of CU, PU, ​​or TU.

[0167] - Depth can have a type depending on the type of partition. QT depth can represent the depth for quadtree partitioning. BT depth can represent the depth for binary partitioning. TT depth can represent the depth for ternary partitioning.

[0168] Sample: A sample can be a base unit that constitutes a block. A sample can consist of one or more bits. Bit depth can be the number of bits that make up a sample. Samples range from 0 to 2 depending on the bit depth. Bd It can be expressed as values ​​up to -1.

[0169] PU: PU may refer to a base unit for processing related to prediction. For example, processing related to prediction may include inter-prediction, intra-prediction, intra-block copy (IBC) prediction, intra-compensation, and motion compensation.

[0170] A single PU can be divided into multiple sub-PUs that are smaller in size than the PU. These multiple sub-PUs can also serve as base units for processing related to prediction. In other words, a prediction unit partition generated by the division of the prediction unit can also be a prediction unit.

[0171] TU: TU may be a base unit for processing related to a residual block. Processing related to a residual block may include at least one of transform, inverse transform, quantization, inverse quantization, transform coefficient encoding, transform coefficient decoding, entropy encoding, and entropy decoding.

[0172] - A single TU can be divided into multiple sub-transformation units that have a size smaller than the size of the TU. Multiple sub-TUs can also serve as base units for processing related to residual blocks. In other words, a transformation unit partition created by the division of a transformation unit can also be a transformation unit.

[0173] - In the embodiments, the sample may include a prediction sample, a restoration sample, a residual sample, and a decoding sample.

[0174] - The transformation may include one or more of a primary transformation and a secondary transformation, and the inverse transformation may include one or more of a primary inverse transformation and a secondary inverse transformation.

[0175] Parameter set: The parameter set can correspond to header information within the structure of the bitstream.

[0176] - The parameter set may include at least one of a Video Parameter Set (VPS), a Sequence Parameter Set (SPS), a Picture Parameter Set (PPS), an Adaptation Parameter Set (APS), and a Decoding Parameter Set (DPS).

[0177] Information signaled through a parameter set can be applied to pictures that reference the parameter set. For example, information within a VPS can be applied to pictures that reference the VPS. Information within an SPS can be applied to pictures that reference the SPS. Information within a PPS can be applied to pictures that reference the PPS. A parameter set can reference a higher-level parameter set. For example, a PPS can reference an SPS. An SPS can reference a VPS.

[0178] - Additionally, the parameter set may include tile group information, slice header information, and tile header information. A tile group may refer to a group or slice containing multiple tiles.

[0179] MPM (Most Probable Mode): MPM may represent an intra-probable mode that is likely to be used for intra-probability of a target block.

[0180] - One or more different MPMs can be determined based on coding parameters related to the target block and attributes of objects related to the target block.

[0181] - One or more MPMs may be determined based on the intra prediction mode of a reference block. There may be multiple reference blocks. One or more different MPMs may be determined depending on which intra prediction modes are used for one or more reference blocks. Reference blocks may include spatial neighbor blocks.

[0182] MPM List: An MPM list may be a list containing one or more MPMs. The number of one or more MPMs in an MPM list may be predefined.

[0183] MPM Index: The MPM index can indicate one or more MPMs in the MPM list that are used for intra prediction for the target block.

[0184] MPM Usage Indicator: The MPM Usage Indicator can indicate whether an MPM list is used for prediction regarding a target block.

[0185] Prediction mode: The prediction mode may be information indicating a prediction method for a target block, such as a mode used for intra-prediction or a mode used for inter-prediction. The prediction mode may refer to one of the prediction-related modes described in the embodiments. Additionally, the prediction mode may include at least one of an intra-mode, an inter-mode, and an intra-block copy mode.

[0186] Reference image list: The reference image list may be a list containing one or more reference images used for prediction of the target block.

[0187] - There may be multiple reference image lists. Multiple reference image lists may include List 0 (List 0; L0), List 1 (List 1; L1), etc.

[0188] - One or more reference image lists may be used for inter prediction for the target block. Parts such as 'L0' and 'L1' in the names of the information related to inter prediction may refer to the reference image lists associated with the information.

[0189] Reference picture: The reference picture may be an image referenced for prediction regarding the target block. Alternatively, the reference picture may be an image containing the reference block. The reference picture may include an image prior to the target image, the target image, and an image following the target image.

[0190] Reference image index: The reference image index may be an index indicating one reference image among one or more reference images in the reference image list that is used for prediction of the target block.

[0191] Reference Block: A reference block may be a block referenced for encoding / decoding of a target block, such as for prediction and filtering. For example, a reference block may include a reference sample referenced to derive a prediction sample, and may refer to a block that provides information used for decoding the target block.

[0192] Reference Sample: A reference sample may be a sample referenced for encoding / decoding of a target block, such as prediction and filtering.

[0193] Inter prediction indicator: The inter prediction indicator may indicate the direction of inter prediction for the target block. Inter prediction may be one of unidirectional prediction and bidirectional prediction. Alternatively, the inter prediction indicator may indicate the number of reference images used when generating prediction blocks for the target block. Alternatively, the inter prediction indicator may indicate the number of prediction blocks used for inter prediction for the target block. The reference direction may refer to the inter prediction indicator. For example, the inter prediction indicator may indicate either unidirectional or bidirectional. Alternatively, for an inter mode that uses only reference images within the L0 reference image list, the inter prediction indicator may have a first value of '0'; for an inter mode that uses only reference images within the L1 reference image list, the inter prediction indicator may have a second value of '1'; and for an inter mode that uses at least two of the reference images within the L0 reference image list and the L1 reference image list, the inter prediction indicator may have a third value of '2'.

[0194] Prediction List Utilization Flag: The prediction list utilization flag for a specific reference image list may indicate whether at least one reference image within that specific reference image list is used to generate the prediction block of the target block. For example, a value of '0' for the prediction list utilization flag for a specific reference image list may indicate that the prediction block is not generated using the reference images within that specific reference image list. A value of '1' for the prediction list utilization flag for a specific reference image list may indicate that the prediction block is generated using the reference images within that specific reference image list.

[0195] - An inter-prediction indicator can be derived using prediction list utilization flags. Conversely, an inter-prediction indicator can be derived using prediction list utilization flags. For example, an inter-prediction indicator can be derived using prediction list utilization flags for multiple reference image lists. If the inter-prediction indicator indicates that specific reference lists among the multiple reference image lists are being used, the prediction list utilization flags of the specific reference lists pointed to by the inter-prediction indicator among the prediction list utilization flags of the multiple reference image lists can be set to '1', and the prediction list utilization flags of the remaining reference image lists not pointed to by the inter-prediction indicator can be set to '0'.

[0196] Reference Direction: The reference direction may point to a list of reference images used for prediction of the target block. For example, the reference direction may point to one or more of reference image list L0 and reference image list L1.

[0197] - The reference direction merely refers to the list of reference images used for prediction of the target block, and does not indicate that the directions of the reference images within the list are restricted to a forward direction or a backward direction. That is to say, each of the reference image list L0 and the reference image list L1 may include forward images and backward images, respectively. Here, the forward direction may indicate a direction from the target image to the image preceding the target image. Forward inter-prediction may be an inter-prediction that uses the image preceding the target image as a reference image. The backward direction may indicate a direction from the target image to the image following the target image. Backward inter-prediction may be an inter-prediction that uses the image following the target image as a reference image.

[0198] - A unidirectional reference direction may mean that a single reference image list is used. A bidirectional reference direction may mean that two reference image lists are used. For example, the reference direction may indicate one of the following: that only reference image list L0 is used, that only reference image list L1 is used, or that two reference image lists are used. Additionally, the reference direction may be indicated by an inter-predictor.

[0199] Picture Order Count (POC): The POC of a picture can represent the display order or output order of the picture.

[0200] Motion information: Motion information may be information used to specify a reference block. Motion information may include information used in a specific prediction mode, such as a motion vector (MV), reference image index, inter prediction indicator, reference picture list information, reference image, prediction list utilization flag, MV candidate, MV candidate index, merge candidate, merge index, block vector, block vector candidate, and block vector candidate index. A specific prediction mode may include an inter prediction mode and an intra block copy mode.

[0201] - Multiple motion information for multiple reference image lists may be used for inter-prediction of the target block. Motion information for a specific reference image list may be used for prediction using that specific reference image list. Multiple (intermediate) prediction blocks may be derived from the multiple motion information. A (final) prediction block for the target block may be generated using statistical values ​​for the multiple (intermediate) prediction blocks.

[0202] MV: MV can be a 2-dimensional vector used in inter-prediction. MV can represent the offset between the target block and the reference block. Alternatively, MV can represent the difference between the location of the target block and the location of the reference block.

[0203] - For example, MV is (mv x , mv y It can be expressed in the form of ). mv x can represent a horizontal component, and mv y It can represent a vertical component.

[0204] - The zero vector can be (0, 0) MV.

[0205] Block Vector (BV): A BV can be a two-dimensional vector used in intra-block copy prediction. A BV can represent the offset between a target block within a target image and a reference block within a target image. In other words, a BV can represent the displacement between a target block and a reference block within a target image.

[0206] - For example, BV is similar to MV (bv x , bv y It can be expressed in the form of ). bv x can represent a horizontal component, and bv y It can represent a vertical component.

[0207] - The zero vector can be (0, 0) BV.

[0208] Motion Information Candidates: In a specific prediction, motion information of the target block can be selected from motion information candidates determined by a specific method. A motion information candidate may refer to the motion information of a reference block, or it may refer to the reference block itself that possesses motion information. Here, the reference block may be a block determined by a specific method to select motion information candidates.

[0209] Candidate List: A candidate list may be a list containing one or more candidates. For example, a candidate list may include a motion information candidate list, a merge candidate list, an MV candidate list, an MPM list, etc. A candidate list may be generated in the same manner in both the encoding device and the decoder. That is to say, the candidate list used in the encoding device and the candidate list used in the decoder may be identical, and the same candidate list may be shared between the encoding device and the decoder. The encoding device may select a candidate from among the candidates in the candidate list to be used for processing the target block. An indicator pointing to the selected candidate may be signaled from the encoding device to the decoder. The decoder may use the indicator to identify the candidate from among the candidates in the candidate list to be used for processing the target block. Alternatively, the encoding device and the decoder may identify the candidate from among the candidates in the candidate list to be used for processing the target block by the same rule.

[0210] Motion Information Candidate List: A motion information candidate list may refer to a list constructed using one or more motion information candidates.

[0211] Motion Information Candidate Index: The motion information candidate index may be an identifier or indicator pointing to a motion information candidate among the motion information candidates in the motion information candidate list that is used for prediction regarding the target block.

[0212] - In a specific inter-prediction mode, motion information of other restored blocks may be used to derive motion information of the target block. Other blocks may include neighboring blocks. In this specific inter-prediction mode, the motion information for the target block itself is not signaled individually, but other information used to derive motion information of the target block based on motion information of other restored blocks may be signaled. In this case, the other information may include information indicating which of the other restored blocks' motion information is used to derive motion information of the target block, such as a motion information candidate index.

[0213] - For example, these inter-prediction modes may include AMVP mode, merge mode, and skip mode. The motion information candidate index may be a merge index or an MV candidate index.

[0214] - In the embodiments, MV may be part of the motion information. In the embodiments, information about motion information, such as motion information candidates, a list of motion information candidates, and an index of motion information candidates, may be replaced with information about MV, such as MV candidates, a list of MV candidates, and an index of MV candidates, and descriptions of motion information may also be applied to MV.

[0215] Merge: Merge can refer to the merging of motion information for multiple blocks, or it can refer to applying the motion information of one block to a target block as well. In other words, merge mode can refer to a mode where the motion information of a target block is derived from the motion information of a neighboring block.

[0216] Merge Candidate: A merge candidate may refer to a specific (restored) block used for merging with a target block, or it may refer to movement information of a specific block. Alternatively, a merge candidate may include movement information of a specific block.

[0217] - Merge candidates for the target block may include spatial merge candidates, temporal merge candidates, history-based candidates, average candidates based on the average of two merge candidates, and zero merge candidates.

[0218] Merge candidate list: The merge candidate list may be a list composed of one or more merge candidates.

[0219] Merge Index: The merge index may be an indicator pointing to a merge candidate among the merge candidates in the merge candidate list that is used for prediction regarding the target block. Among the merge candidates in the merge candidate list, the movement information of the merge candidate indicated by the merge index may be used as movement information for the target block.

[0220] Neighbor block: A neighbor block may refer to a block adjacent to the target block. Neighbor blocks may include spatial neighbor blocks and temporal neighbor blocks. A neighbor block may also refer to a reconstructed neighbor block within the reference image.

[0221] - The neighbor blocks of the target block may refer to the blocks adjacent to the target block.

[0222] - Neighbor blocks do not necessarily have to be in direct contact with the target block. The neighbor blocks of the target block may be blocks whose distance from the target block is less than or equal to a specific value.

[0223] Distance: In the embodiments, the distance between objects may be used. Each object may be a block, line, pixel, or template, etc.

[0224] - The distance between objects may be the larger of the horizontal distance and the vertical distance. Or, the distance may be the horizontal distance; the vertical distance; the diagonal distance; or the smaller of the horizontal distance and the vertical distance.

[0225] - The distance may be the shortest distance between objects, and may be the distance between specific locations described in the embodiments within the objects.

[0226] - The unit of distance may be a pixel or a block. Additionally, the unit of distance may be a specific unit described in one of the objects or embodiments.

[0227] Spatial neighbor blocks: Spatial neighbor blocks can be blocks that are spatially adjacent to the target block.

[0228] - The target block and spatial neighbor blocks can be included within the target image.

[0229] - Spatial neighbor blocks may include blocks whose boundaries, at least a portion of which abuts at least a portion of the target block's boundary. Alternatively, spatial neighbor blocks may include blocks whose distance from the target block is less than or equal to a specific value.

[0230] - Spatial neighbor blocks may include blocks diagonally adjacent to the vertices of the target block.

[0231] - Spatial neighbor blocks may include a top-left block adjacent to the top-left of the target block, a top block adjacent to the top of the target block, a top-right block adjacent to the top-right of the target block, a left block adjacent to the left of the target block, a right block adjacent to the right of the target block, a bottom-left block adjacent to the bottom of the target block, and a bottom-right block adjacent to the bottom-right of the target block.

[0232] Temporal neighbor blocks: Temporal neighbor blocks can be blocks that are temporally adjacent to the target block.

[0233] - Temporal neighbor blocks may include a collocated block (COL block). A collocated block may be a block within a restored image in a reference image buffer. A collocated picture (col picture) may refer to an image containing a collocated block. A collocated picture may be an image included in a reference image list.

[0234] - Call blocks can be determined based on the location of target blocks within the target image. Two blocks being 'temporarily adjacent' may mean that the locations of the two blocks satisfy certain conditions.

[0235] - The position of the call block within the call image may be the same as the position of the target block within the target image. Alternatively, the position of the call block within the call image may correspond to the position of the target block within the target image. Here, the correspondence of the block positions may mean that the regions of the blocks are identical, that the region of one block is included within the region of another block, or that one block occupies a specific location within another block.

[0236] - For example, the location of a call block within a call image may be the same as the location of a target block within a target image. Alternatively, the call block may be a block containing call samples within a call image. A call sample may be a sample having coordinates identical to the coordinates of a specific sample in the target block.

[0237] - Temporal neighbor blocks may be blocks that are temporally adjacent to the spatial neighbor blocks of the target block.

[0238] Search range: The search range may refer to a two-dimensional area where a search for an MV is performed during inter-prediction. For example, when an optimal MV needs to be derived for processing a target block, the optimal MV can be selected from among the MVs pointing inside the search range.

[0239] Transform coefficient: The transform coefficient may be a coefficient generated by performing a transformation on the residual block. Alternatively, the transform coefficient may be a coefficient value generated by performing inverse quantization on the quantized level.

[0240] Quantized level: A quantized level can be an integer quantity used as an input for inverse quantization.

[0241] Quantization: Quantization can be a process that generates quantized levels for transformation coefficients. Quantized levels can be generated by applying quantization to transformation coefficients. Transformation can also be considered as part of quantization.

[0242] Inverse Quantization: Inverse quantization can be a process of multiplying a quantized level by a factor. By applying inverse quantization to the quantized level, (restored) transformation coefficients can be generated.

[0243] Quantization Parameter (QP): QP may refer to the argument used to generate quantized levels for transform coefficients in quantization. Additionally, QP may refer to the argument used to generate (restored) transform coefficients for quantized levels in inverse quantization. Alternatively, QP may be a value mapped to the quantization step size.

[0244] Delta QP: Delta QP can be the difference between the QP predicted by a specific process and the QP of the target block. In other words, the QP of the target block can be the sum of the predicted QP and Delta QP.

[0245] Quantization matrix: A quantization matrix may be a matrix used in quantization or inverse quantization to improve the subjective or objective image quality.

[0246] Quantization matrix coefficients: Quantization matrix coefficients can be each element within the quantization matrix.

[0247] Scan: Scan can refer to a method of arranging values ​​within a block or matrix. The values ​​can be coefficients. For example, a scan can mean arranging values ​​arranged in a 2D form into a 1D form, or rearranging values ​​arranged in a 1D form into a 2D form. An inverse scan can be the opposite arrangement (or rearrangement) of the arrangement performed in a scan.

[0248] Non-zero transformation coefficients: Non-zero transformation coefficients may refer to transformation coefficients that have a non-zero value or quantized levels that have a non-zero value.

[0249] Bitstream: A bitstream may refer to a series or sequence of bits containing encoded information generated by encoding of an image. A bitstream may contain information according to specific syntax elements. For example, the information may include syntax elements. An encoding device may generate a bitstream containing information according to specific syntax elements. A decoder may obtain information from the bitstream according to specific syntax elements.

[0250] Signaling: Signaling of information may indicate that information is transmitted from an encoding device to a decoding device via a bitstream. For example, the information may include syntactic elements. Alternatively, signaling may mean that the encoding device includes information within the bitstream. Information signaled by the encoding device may be used by the decoding device. In signaling, the bitstream may be transmitted over a network and may be contained within a recording medium. In embodiments, the description that information is signaled may include: 1) the encoding device determining and generating information for signaling of information; 2) the encoding device performing encoding on the information to generate encoded information; 3) the (encoded) information being transmitted from the encoding device to the decoding device via a bitstream; 4) the decoding device performing decoding on the encoded information to obtain information; and 5) the decoding device determining and generating information through signaling of information.

[0251] - An encoding device can generate encoded information by performing encoding on the information. The encoded information can be signaled through a bitstream. A decoding device can obtain information by performing decoding on the encoded information.

[0252] - The fact that information is signaled to a specific target may mean that the information is used for each specific target, and that the processing represented by the information is applied to each specific target. For example, the fact that information is signaled at a specific unit level may indicate that the information is used or processed for each specific unit.

[0253] - The signaled information may include one or more sub-information. That specific information is signaled may mean that each piece of information of the one or more sub-information included in the specific information is signaled.

[0254] Optional Signaling: Signaling for information may be performed optionally. Optional signaling for information may mean that an encoding device optionally includes information within a bitstream (depending on specific conditions). Optional signaling for information may mean that a decoder optionally obtains information from a bitstream (depending on specific conditions).

[0255] Omission of Signaling: Signaling for information may be omitted. Omission of signaling for information may mean that the encoding device does not include information in the bitstream (depending on specific conditions). Omission of signaling for information may mean that the decoding device does not obtain information from the bitstream (depending on specific conditions). The decoding device may derive information with omitted signaling using other information of the embodiments.

[0256] Symbol: May represent at least one piece of information of a target unit, such as syntactic elements, coding parameters, quantized levels, and transform coefficients of a target unit or target block. Additionally, the symbol may represent the target of entropy encoding or the result of entropy decoding.

[0257] Entropy encoding: Entropy encoding can allocate a small number of bits to symbols with a high probability of occurrence and a large number of bits to symbols with a low probability of occurrence. Through this allocation, the size of the bitstream representing the symbols can be reduced.

[0258] Entropy coding can utilize methods such as Variable Length Coding (VLC) and Context-Adaptive Binary Arithmetic Coding (CABAC). For example, in Variable Length Coding, entropy coding can be performed using variable-length tables. For instance, in CABAC, a binaryization method for symbols and a probabilistic model of symbols / bins can be derived for entropy coding, and context-based arithmetic coding can be performed.

[0259] Entropy Decoding: In entropy decoding, the processes performed in entropy encoding can be performed in reverse. Symbols can be generated by entropy decoding of a bitstream.

[0260] Parsing: Parsing can refer to determining the values ​​of syntactic elements by performing entropy decoding on the encoded information of a bitstream. Alternatively, parsing can refer to entropy decoding itself.

[0261] Statistical Value: The values ​​of information related to specific entity(s) described in the embodiments may be used as inputs for specific operations. The statistical value may be a value derived by a specific operation on the values ​​related to these specific entity(s). For example, the statistical value for specific information may be one or more of the following: an average value, a weighted average value (weighted average), a weighted sum (weighted sum), a minimum value, a maximum value, a mode, a median value, an interpolated value, a sum of products, and a product of sums. Additionally, information of the embodiment having specific values ​​determined by operations, such as constants, variables, and coding parameters, may have a specific statistical value according to the embodiment.

[0262]

[0263] Coding parameters

[0264] In the embodiments, the coding parameters may be information required for coding. The coding parameters may include information signaled from an encoding device to a decoder, information calculated / derived during the processing of coding described in the embodiments, and information used for the processing of coding described in the embodiments.

[0265] In the embodiments, the coding parameters include the size of the CTU, the size of the unit, the form of the unit, the shape of the unit, the depth of the unit, the minimum unit size, the maximum unit size, the maximum unit depth, the minimum unit depth, the unit splitting information, QT splitting information, BT splitting information, the splitting direction of the BT splitting, the splitting form of the BT splitting, TT splitting information, the splitting direction of the TT splitting, the splitting form of the TT splitting, MTT splitting information, the combination of MTT splittings, the splitting direction of the MTT splitting, the splitting form of the MTT splitting, the prediction mode, the intra prediction mode, the luminance intra prediction mode, the chroma intra prediction mode, the intra prediction mode, the inter splitting information, the coding block splitting information, the prediction block splitting information, the transformation block splitting information, the reference sample line index, the reference sample filtering method, the reference sample filter tab, the reference sample filter coefficients, the prediction block filtering method, the prediction block filter tab, the prediction block filter coefficients, the prediction block boundary filtering method, the prediction block boundary filter tab, the prediction block boundary filter coefficients, the inter prediction mode, motion information, MV, and Motion Vector Difference; MVD), MVD resolution, MV size, MV representation accuracy, reference image list, reference image, reference image index, inter prediction direction, inter prediction indicator, prediction list utilization flag, POC, MV candidate, MV candidate index, MV candidate list, AMVP mode usage information, merge candidate, merge index, merge candidate list, merge mode usage information, motion information correction information, skip mode usage information, intra-block copy mode usage information, BV (Block Vector), Block Vector Difference (BVD), BVD resolution, BV size, BV representation accuracy, BV candidate, BV candidate index, BV candidate list, interpolation filter filter tab, interpolation filter filter coefficients, transform type, transform size, transform selection information, primary transform usage information,Secondary transform usage information, primary transform selection information, secondary transform selection information, residual block presence information, coded block pattern, coded block flag, QP, delta QP, quantization matrix, deblocking filter usage information, deblocking filter coefficients, deblocking filter filter tab, deblocking filter strength, deblocking filter shape / form, adaptive sample offset usage information, adaptive sample offset value, adaptive sample offset category, adaptive sample offset type, adaptive loop filter usage information, adaptive loop filter coefficients, adaptive loop filter filter tab, adaptive loop filter shape / form, binarization / debinarization method, context model, context model determination method, context model update method, regular mode usage information, bypass mode usage information, significant coefficient flag, last significant coefficient flag, coefficient group coding flag, last significant coefficient position, flag indicating whether the coefficient value is greater than 1, whether the coefficient value is greater than 2 Flag indicating presence, flag indicating whether the coefficient value is greater than 3, remaining coefficient value information, sign information, context bin, bypass bin, restored sample, restored luminance sample, restored chroma sample, residual sample, residual luminance sample, residual chroma sample, transform coefficient, luminance transform coefficient, chroma transform coefficient, transform coefficient level, luminance transform coefficient level, chroma transform coefficient level, transform coefficient level scanning method, quantized level, luminance quantized level, chroma quantized level, size of the MV seek area on the decoder side, shape of the MV seek area on the decoder side, number of MV seeks on the decoder side, picture type, slice identification information, slice type, slice splitting information, tile group identification information, tile group type, tile group splitting information, tile identification information, tile type, tile splitting information, bit depth,It may include one or more of input sample bit depth, restored sample bit depth, residual sample bit depth, transform factor bit depth, quantized level bit depth, mapping availability information, information about the luminance signal, information about the chroma signal, the color space of the target block, the color space of the residual block, and temporal layer information.

[0266] In addition, the coding parameter may further include 1) a value of information that may be included in the coding parameter, 2) a combination of multiple pieces of information that may be included in the coding parameter, 3) a statistical value of information that may be included in the coding parameter, 4) information related to the coding parameter, 5) information used to calculate / derive the coding parameter, and 6) information calculated / derived using the coding parameter.

[0267] In the embodiments, "X usage information" may be "information indicating whether X is used / applied / executed." Alternatively, "X usage information" may be "information indicating whether X is available." For example, "specific mode usage information" may be information indicating whether a specific mode is used. Mode information may indicate a mode used for a target block among the modes described in the embodiments. In the embodiments, specific mode usage information may be replaced with mode information, and the description of specific mode usage information may also apply to mode information. "X usage information" and "X indicator" may be used interchangeably.

[0268] In the embodiments, coding parameters and syntax elements may correspond to each other. For example, a syntax element of the embodiment may be used as a coding parameter, and a coding parameter may be signaled as a syntax element.

[0269] In the embodiments, "X existence information" may be considered as "information indicating whether X exists" or "information indicating whether information indicating X exists within the bitstream".

[0270] In the embodiments, "X selection information" may be information indicating one of the candidates or methods for X. "X selection information" may be considered as an "X index".

[0271] In the embodiments, the splitting form of a specific tree may represent one of symmetric splitting and asymmetric splitting, and may represent one of QT, BT, TT, and non-split. The splitting direction of a specific tree may represent one of horizontal direction and vertical direction.

[0272] In the embodiments, when the coding parameter has one of a plurality of values, "coding parameter" may be replaced with "whether the coding parameter has a specific value among the plurality of values ​​available to the coding parameter".

[0273] In the embodiments, when the coding parameter refers to one of a plurality of targets, the "coding parameter" may be replaced with "whether the coding parameter refers to a specific target among the plurality of targets."

[0274]

[0275] indicator

[0276] In the embodiments, the expressions "method," "mode," and "method using the mode (encoding / decoding)" may be used interchangeably.

[0277] Whether a specific method is performed may be indicated by an indicator for the specific method. The "indicator" for the "specific method" may be abbreviated as "specific indicator".

[0278] In the embodiments, "when an indicator indicating whether a specific method is performed is true" may mean "when it is true whether the specific method is performed on the object indicated by the indicator; or on the object associated with the indicator."

[0279] In the embodiments, "where a specific object is a specific method" may mean a case where the specific method is performed on the specific object.

[0280] In the embodiments, the target indicated by the indicator may include the unit described in the embodiments; the sample described in the embodiments; the location of the unit or the sample; the mode described in the embodiments, such as a prediction mode; the information described in the embodiments, such as motion information; and coding parameters.

[0281] For example, an indicator indicating whether a specific method is performed may have one of a plurality of values, and the specific method may be performed only when the indicator has one of some of the aforementioned plurality of values. Here, "when the indicator indicating whether a specific method is performed is true" may mean the case where the indicator indicating whether a specific method is performed has one of the aforementioned partial values.

[0282] In the embodiments, "when the indicator for a specific judgment is true" may mean that the result of the specific judgment is true in the object indicated by the indicator; or in the object associated with the indicator.

[0283] In the embodiments, "when a specific method is performed" may mean that an indicator indicating whether a specific method is performed has a true value.

[0284] In the embodiments, if a specific method is not activated at a specific target, at least one signaling / encoding / decoding of syntactic elements for the specific target may be omitted at the specific target and the sub-targets of the specific target.

[0285] In the embodiments, the fact that a specific method is performed only under specific conditions may mean that the specific mode is enabled only when the specific conditions are met.

[0286]

[0287] System for video coding

[0288] FIG. 1 shows a system for video coding according to one embodiment.

[0289] The system (100) may include at least one of an encoding device (110) and a decoding device (150).

[0290] Each of the encoding device (110) and the decoding device (150) may be a computer or an electronic apparatus.

[0291]

[0292] Structure of the encoding device

[0293] The encoding device (110) may include a processor (120), a storage (140), and a communicator (149).

[0294] The processor (120), storage (140), and communication device (149) can be connected via a bus.

[0295] The processor (120) may be a semiconductor device that executes instructions or computer-executable code, such as a Central Processing Unit (CPU). The processor (120) may be at least one hardware processor.

[0296] The processor (120) can perform generation and processing of information that is input to the encoding device (110) in the embodiments, output from the encoding device (110), or used inside the encoding device (110), and can perform comparison and judgment related to such information.

[0297] The processor (120) may include a plurality of components. The plurality of components may include a partitioner (122), a subtractor (124), a transformer (125), a quantizer (126), an inverse quantizer (127), an inverse transformer (128), an adder (129), a filter (130), and an entropy encoder (139).

[0298] At least some of the aforementioned multiple components may be program modules. Program modules may be included in the encoding device (110) in the form of an operating system, an application, and other program modules. Program modules may be instructions or computer-executable code stored in a storage (140) and executed by a processor (120).

[0299] The storage (140) may include various types of volatile storage media and non-volatile storage media. For example, the storage (140) may include memory such as ROM and RAM.

[0300] The storage (140) can store instructions and computer-executable code used for the operation of the encoding device (110), and can store information and bitstreams as described in the embodiments. The storage (140) may include a reference picture buffer (141).

[0301] The communication device (149) can perform functions related to the communication of information in the encoding device (110). For example, the communication device (149) can transmit a bitstream to the decoding device (150).

[0302] Among the names of the components of the encoding device (110), "-gi" ("-er" or "-or") may be replaced with "-bu" (- unit). The storage unit (140) may also be named a storage unit.

[0303]

[0304] Operation of the encoding device

[0305] The encoding device (110) can sequentially encode one or more images of the video.

[0306] The storage (140) can store the original image. In the encoding device (110), the original image can be used as the target image.

[0307] The processor (120) can generate a bitstream containing encoded information by performing encoding on the target image and can store the generated bitstream in a storage (140). The generated bitstream can be stored in a computer-readable recording medium and can be transmitted by the communication device (149) to the communication device (189) of the decoding device (150) via a wired and / or wireless transmission medium.

[0308] The splitter (122) can determine the target block by performing a split on the target image.

[0309] The predictor (123) can determine the prediction mode of the target block. The predictor (123) can generate a prediction block of the target block by performing a prediction according to the prediction mode.

[0310] The prediction mode of the target block may be one of the available prediction modes. For example, available prediction modes may include intra prediction, inter prediction, and IBC prediction.

[0311] For example, if the prediction mode is intra prediction, the predictor (123) can perform intra prediction on the target block to generate a prediction block of the target block.

[0312] For example, if the prediction mode is inter-prediction, the predictor (123) can perform inter-prediction on the target block to generate a prediction block of the target block.

[0313] For example, if the prediction mode is IBC, the predictor (123) can perform an IBC prediction for the target block to generate a prediction block of the target block.

[0314] The subtractor (124) can generate a residual block of the target block. The residual block may be the difference between the original block and the prediction block. The original block may be the region of the original image pointed to by the target block. Alternatively, the residual block may refer to a block generated by applying one or more of transformation and quantization to the difference between the original block and the prediction block.

[0315] The converter (125) can perform a conversion on the residual block to generate conversion coefficients.

[0316] The converter (125) can perform the conversion using one of a plurality of conversion methods.

[0317] For example, multiple transformation methods may include the Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Karhunen-Loeve Transform (KLT), and transformations based on each transformation.

[0318] The transform skip mode may be a mode that generates a restored block using the restored residual block and prediction block, for which transform and inverse transform have not been performed. When the transform skip mode is applied to a target block, the transform and inverse transform for the target block may be omitted, and only quantization and inverse quantization for the target block may be performed.

[0319] The quantizer (126) can generate quantized levels by applying quantization using quantization parameters to the conversion coefficients. In the embodiments, the quantized levels may also be referred to as conversion coefficients.

[0320] The entropy encoder (139) can generate encoded information by performing entropy encoding based on a probability distribution on information for decoding an image. The bitstream may contain encoded information.

[0321] Information for decoding the image may include quantized levels and syntax elements produced by the quantizer (126).

[0322] The probability distribution can be determined based on quantized levels and coding parameters.

[0323] The entropy encoder (139) can convert quantized levels, which have the form of a two-dimensional block, into the form of a one-dimensional vector by using scanning to perform encoding for the quantized levels. In the scanning, it can be determined which scan to use among an upper-right diagonal scan, a vertical scan, and a horizontal scan based on coding parameters such as the size of the block and the intra-prediction mode of the block.

[0324] When encoding is performed on a target image / block, the predictor (123) uses a reference image / block for prediction. The encoded target image / block can be used as a reference image / block for other images / blocks that are subsequently processed. Accordingly, the processor (120) can perform restoration on the encoded target block and can store the restored image containing the restored target block generated by the restoration in the reference picture buffer (141) as a reference image. Inverse quantization and inverse transform can be performed on the encoded target block for restoration.

[0325] The inverse quantizer (127) can generate inverse quantized conversion coefficients by performing inverse quantization on the quantized level.

[0326] The inverse converter (128) can generate inversely quantized and inversely converted coefficients by performing an inverse conversion on the inversely quantized conversion coefficients. In embodiments, the inversely quantized and / or inversely converted coefficients may refer to coefficients to which at least one of the inverse quantization and inverse conversion has been applied. The inversely quantized and inversely converted coefficients may be restored residual blocks.

[0327] The adder (129) can generate a recovery block by combining the prediction block and the recovered residual block.

[0328] The restoration block may pass through a filter (130). The filter (130) may apply one or more of a plurality of filters to the target. Each of the plurality of filters may be an in-loop filter. The target may be a restoration sample, a restoration block, or a restoration image.

[0329] The reference picture buffer (141) can store a restoration block / image provided from the filter (130). The restoration image may be an image containing the restoration block. Alternatively, the restoration image may be an image composed of restoration blocks.

[0330] The reference picture buffer (141) can provide the stored restored image to the predictor (123) as a reference image. In terms of storing the decoded (i.e., restored) picture, the reference picture buffer (141) may also be referred to as the Decoded Picture Buffer (DPB).

[0331]

[0332] Structure of the decoding device

[0333] The decoding device (150) may include a processor (160), a storage device (180), and a communication device (189).

[0334] The description of the processor (120), storage (140), and communication device (149) associated with the encoding device (110) may also apply to the processor (160), storage (180), and communication device (189) associated with the decoding device (150). Redundant descriptions are omitted.

[0335] The processor (160) may include a plurality of components. The plurality of components may include an entropy decoder (161), a splitter (162), a predictor (163), an inverse quantizer (167), an inverse converter (168), an adder (169), and a filter (170).

[0336] The storage (180) may include a reference picture buffer (181).

[0337] The communicator (189) can perform functions related to the communication of information in the decoding device (150). For example, the communicator (189) can receive a bitstream from the encoding device (110).

[0338] Among the names of the components of the decoding device (150), "-gi" ("-er" or "-or") may be replaced with "-bu" (- unit). The storage unit (180) may also be named a storage unit.

[0339]

[0340] Operation of the decoding device

[0341] A communication device (149) of an encoding device (110) can transmit a bitstream generated by the encoding device (110) to a decoding device (150). Alternatively, a computer-readable recording medium storing the bitstream can transmit the bitstream generated by the encoding device (110) to a decoding device (150).

[0342] The communication device (189) can receive a bitstream from the encoding device (110) via a wired and / or wireless transmission medium. The received bitstream can be stored in a storage device (180).

[0343] The processor (160) can obtain a bitstream from a storage (180) or a computer-readable recording medium.

[0344] A bitstream can contain encoded information.

[0345] The entropy decoder (161) can generate information for decoding an image by performing entropy decoding based on a probability distribution on the encoded information of the bitstream.

[0346] Information for decoding an image may include quantized levels and syntax elements, etc.

[0347] The entropy decoder (161) can convert quantized levels, which have the form of a one-dimensional vector, into the form of a two-dimensional block by using scanning to perform decoding on the quantized levels. In the scanning, it can be determined which scan to use among an upper-right diagonal scan, a vertical scan, and a horizontal scan based on coding parameters such as the size of the block and the intra-prediction mode of the block.

[0348] The entropy decoder (161) can provide syntax elements to other components of the processor (160), such as the splitter (162).

[0349]

[0350] Common explanation based on the relationship between the components of the encoding device and the components of the decoding device

[0351] The decoding device (150) performs decoding using the bitstream generated by the encoding device (110). The encoding device (110) may perform encoding for the target block using a restored image derived within the decoding device (150), rather than an original image that is not provided to the decoding device (150). Accordingly, the encoding device (110) and the decoding device (150) may need to generate the restored block / image in the same way. In this regard, the descriptions of the divider (122), predictor (123), inverse quantizer (127), inverse converter (128), adder (129), filter (130), and reference picture buffer (141) of the encoding device (110) disclosed in the embodiments may also be applied to the divider (162), predictor (163), inverse quantizer (167), inverse converter (168), adder (169), filter (170), and reference picture buffer (181) of the decoding device (150), respectively. Redundant descriptions are omitted.

[0352] Additionally, each of the divider (122), predictor (123), inverse quantizer (127), inverse converter (128), adder (129), and filter (130) of the encoding device (110) can generate syntactic element information that specifies processing for a target. Each of the divider (162), predictor (163), inverse quantizer (167), inverse converter (168), adder (169), and filter (170) of the decoding device (150) can perform processing for a target (such as that performed in the encoding device (110)) using the syntactic element information.

[0353] As described above, corresponding components of the encoding device (110) and the decoding device (150) may perform the same or corresponding functions. In embodiments, the processor may represent the processor (120) of the encoding device (110) and / or the processor (160) of the decoding device (150). For example, regarding the function of prediction, the processor may represent a predictor (123), a subtractor (124), and an adder (129), and may represent a predictor (163) and an adder (169). Regarding the function of conversion, the processor may represent a converter (125) and an inverse converter (128), and may represent an inverse converter (168). Regarding the function of quantization, the processor may represent a quantizer (126) and an inverse quantizer (127), and may represent an inverse quantizer (167). In terms of functions related to entropy encoding / decoding, the processing unit may represent an entropy encoder (139) and / or an entropy decoder (161). In terms of functions related to filtering, the processing unit may represent a filter (130) and / or a filter (170). The storage unit may represent a storage unit (140) of the encoding device (110) and / or a storage unit (180) of the decoding device (150). The reference picture buffer may represent a reference picture buffer (141) of the encoding device (110) and / or a reference picture buffer (181) of the decoding device (150). The communication unit may represent a communication unit (149) of the encoding device (110) and / or a communication unit (189) of the decoding device (150).

[0354]

[0355] Partitioning of the units that constitute the image

[0356] Figure 2 shows a segmentation structure of an image according to one embodiment.

[0357] Figure 2 schematically illustrates an example in which a single unit is divided into multiple sub-units.

[0358] CU can be used as a base unit for encoding and decoding of images. Additionally, CU can be a base unit for prediction, transformation, quantization, inverse quantization, inverse transformation, entropy encoding, and entropy decoding.

[0359] A CU can be used as a unit to which a prediction mode is applied. That is to say, in coding, it can be determined which of the available prediction modes will be applied to each CU. For example, available prediction modes may include intra prediction, inter prediction, and intra block copy (IBC) prediction.

[0360] The target image (200) can be sequentially divided into units of CTUs. A division structure can be determined for each CTU. The CTU can be divided into CUs according to the division structure. Alternatively, one CTU can be used as a CU. The size of the CTU can be the maximum size of the CU.

[0361] Each CU may have depth information. The depth information may represent the depth of the CU and the size of the CU. The depth of the CTU may be 0. The depth of the CU created by dividing the CTU may be 1. When a parent CU is divided into child CUs, the depth of the child CU may be 1 greater than the depth of the parent CU. The number of divided CUs may be a positive integer greater than or equal to 2, including 2, 4, 8, and 16. At least one of the width and height of the child CU created by dividing the parent CU may be smaller than at least one of the width and height of the parent CU, depending on the number of child CUs.

[0362] A partitioned CU can be recursively partitioned in the same way up to a predefined maximum depth or a predefined minimum size. The depth of a Smallest Coding Unit (SCU) can be the predefined maximum depth, and the size of an SCU can be the predefined minimum size. The size of an SCU can be the minimum CU size.

[0363] For example, the depth range of a CU can be values ​​from 0 to 3. Depending on the depth of the CU, the CU can have a size from 64x64 to 8x8. A CTU with a depth of 0 can be 64x64 blocks. 0 can be the minimum depth. An SCU with a depth of 3 can be 8x8 blocks. 3 can be the maximum depth. Depth 0 can represent a CTU that is 64x64 blocks. Depth 1 can represent a CU that is 32x32 blocks. Depth 2 can represent a CU that is 16x16 blocks. Depth 3 can represent an SCU that is 8x8 blocks.

[0364] The partition information of a CU may indicate whether the CU is partitioned. The partition information may be a 1-bit flag. All CUs except the SCU may include partition information. For example, the partition information of a CU that is not further partitioned may be a first value of '0', and the partition information of a CU that is partitioned may be a second value of '1'.

[0365] Quad Tree (QT) partitioning can mean that a single CU is partitioned into four CUs. When a parent CU is partitioned into four child CUs, the width and height of each child CU can be half the width and half the height of the parent CU, respectively.

[0366] A binary tree (BT) partition can mean that one CU is divided into two CUs. For example, if a parent CU is divided into two child CUs, the width or height of each child CU can be half the width or half the height of the parent CU.

[0367] Ternary tree (TT) partitioning can mean that a single CU is divided into three CUs. For example, when a parent CU is divided into three child CUs, the three child CUs can be created by dividing the width or height of the parent CU in a ratio of 1:2:1. The width or height of the child CUs can be 1 / 4, 1 / 2, and 1 / 4 of the width or height of the parent CU, respectively.

[0368] In FIG. 2, QT-type splitting was applied to the first CTU. QT splitting, BT splitting, and TT splitting were applied to the second CTU.

[0369] To split a CTU, at least one of different types of splits, such as QT splitting, BT splitting, and TT splitting, may be applied to the CTU. Different types of splits may be applied based on specific priorities.

[0370] For example, QT splitting may be applied preferentially to a CTU. A CU to which QT splitting can no longer be applied may correspond to a leaf node of QT. A CU that is a leaf node of QT may become a root node of BT and / or TT. A CU that is a leaf node of QT may be split into a BT form or a TT form, or may not be split further. In this case, QT splitting may not be applied again to a CU created by applying BT splitting or TT splitting to a CU that is a leaf node of QT.

[0371] The splitting of a CU corresponding to each node of QT can be signaled using QT splitting information. The QT splitting information may be a flag. The QT splitting information of a unit may be information indicating whether the unit is split into a QT form. A first value of the QT splitting information, '0', may indicate that the CU is not split into a QT form. QT splitting information having a first value may signify a Multi-Type Tree (MTT) split. MTT splitting may include BT splitting and TT splitting. A second value of the QT splitting information, '1', may indicate that the CU is split into a QT form.

[0372] There may be no priority between BT splitting and TT splitting. That is, CUs corresponding to the leaf nodes of QT can be split into BT form or TT form. Additionally, CUs generated by BT splitting or TT splitting can be split again into BT form or TT form, or they may not be split any further.

[0373] A CU corresponding to a leaf node of QT can be a root node of MTT. For a CU corresponding to each node of MTT, the CU may further include partition direction information and partition type information in the form of MTT.

[0374] The splitting direction information can indicate the splitting direction of the MTT split. The first value of the splitting direction information, '0', can indicate that the CU is split in the horizontal direction. The second value of the splitting direction information, '1', can indicate that the CU is split in the vertical direction.

[0375] The split type information may indicate the split type used for multi-type tree splitting. The first value of the split type information, '0', may indicate that the CU is split into a TT type. The second value of the split type information, '1', may indicate that the CU is split into a BT type.

[0376] Here, each of the aforementioned division direction information and division shape information may be a flag having a specific length (e.g., 1 bit).

[0377] The partitioning information of CU may also include QT partitioning information, partitioning direction information, and partitioning shape information.

[0378] A CU that is no longer divided by QT division, BT division, and TT division can be used as a unit for specific processing such as prediction, transformation, quantization, inverse quantization, inverse transformation, entropy encoding, and entropy decoding. That is, for a specific processing, the CU may no longer be divided. Therefore, division information for dividing such a CU into PU and / or TU, etc., may not exist within the bitstream.

[0379] On the other hand, if the size of a CU is larger than the maximum TU size, such a CU can be recursively partitioned until the size of the CU becomes less than or equal to the maximum TU size. For example, if the size of the CU is 64x64 and the maximum TU size is 32x32, the CU can be partitioned into 4 32x32 TUs for transformation. For example, if the size of the CU is 32x64 and the maximum TU size is 32x32, the CU can be partitioned into 2 32x32 TUs for transformation.

[0380] In such cases, information regarding whether the CU is split for transformation may not be signaled separately. Whether the CU is split may be determined without signaling by comparing the size of the CU (width / height) and the maximum TU size (width / height). For example, if the width of the CU is greater than the width of the maximum TU size, the CU may be split vertically into two. Additionally, if the height of the CU is greater than the height of the maximum TU size, the CU may be split horizontally into two.

[0381] For example, the minimum size of a CU can be 4x4. For example, the maximum size of a transformation block can be 64x64. For example, the minimum size of a transformation block can be 4x4. The minimum size of QT can be the minimum size of a CU corresponding to a leaf node of QT. The maximum depth of MTT can be the maximum depth of a path from the root node of MTT to a leaf node.

[0382] The BT maximum size may represent the maximum size of the CU corresponding to each node of the BT, and the TT maximum size may represent the maximum size of the CU corresponding to each node of the TT. The BT minimum size and / or the TT minimum size may be set as the minimum size of the CU.

[0383] If the depth of a CU within the MTT corresponding to a node of the MTT is equal to the maximum depth of the MTT, the CU may not be divided into BT form and / or TT form.

[0384] Based on the various sizes and depths of the aforementioned CU, each piece of information described in the embodiments may or may not be present in the bitstream.

[0385] Information regarding the maximum or minimum size described in the embodiments may be signaled at the upper level of the CU. In the embodiments, the upper level of the CU may include a video level, a sequence level, a picture level, a subpicture level, a tile group level, a tile level, and a slice level, etc.

[0386] The information described in the embodiments may be signaled separately for different types of slices. Different types of slices may include intra-slices and inter-slices.

[0387]

[0388] Processing of blocks based on block attributes

[0389] Whether a specific process described in the embodiments is applied or performed may be determined based on the attributes of the block associated with the specific process. Whether a specific process described in the embodiments is applied or performed may be determined based on whether the attributes of the block associated with the specific process satisfy specific conditions. For example, a block may include a target block, a neighbor block, and a reference block. A block may include other blocks described in the embodiments. A block may be one of the blocks and units described in the embodiments.

[0390] The block to which the specific treatment described in the embodiments is applied may have a square shape or a non-square shape.

[0391] In one embodiment, the attributes of the block may include the size of the block. The specific processing described in the embodiments may be applied / performed when specific conditions regarding the size of the block are met.

[0392] In one embodiment, specific conditions may include a minimum block size condition and a maximum block size condition. The block to which the minimum block size condition applies and the block to which the maximum block size condition applies may be different from each other.

[0393] In one embodiment, the minimum block size and / or maximum block size for a specific process may be predefined.

[0394] In one embodiment, the processing of the embodiment may be applied / performed when the block size is greater than or equal to the minimum block size and / or less than or equal to the maximum block size. Alternatively, in one embodiment, the processing of the embodiment may be applied / performed when the block size is greater than the minimum block size and / or less than the maximum block size.

[0395] In one embodiment, the processing of the embodiment may be applied / performed only when the block size is greater than or equal to the minimum block size and less than or equal to the maximum block size. Alternatively, the processing of the embodiment may be applied / performed only when the block size is greater than the minimum block size and less than or equal to the maximum block size. Alternatively, the processing of the embodiment may be applied / performed only when the block size is greater than the minimum block size and less than the maximum block size. The processing of the embodiment may be applied / performed only when the block size is greater than the minimum block size and less than the maximum block size.

[0396] In one embodiment, the processing of the embodiment may be applied / performed only when the block size is a predefined block size.

[0397] In the embodiments, the size of the block may be determined by various methods. For example, the size of the block may mean the width of the block or the height of the block. The size of the block may mean both the width and the height of the block. The size of the block may mean the area of ​​the block. The size of the block may mean 1) the result of a known formula using the width and height of the block, 2) the result of a formula of the embodiment, or 3) a statistical value.

[0398] Additionally, for the first size, the treatment of the first embodiment among the embodiments may be applied / performed, and for the second size, the treatment of the second embodiment among the embodiments may be applied / performed.

[0399] In the embodiments, the block size may be 2x2, 4x4, 8x8, 16x16, 32x32, 64x64, or 128x128, etc. Or, in the embodiments, the block size is (2*SIZE X )x(2*SIZE Y It may be ) etc. SIZE X is one of integers greater than or equal to 1. SIZE Y can be one of integers greater than or equal to 1.

[0400]

[0401] Prediction information for prediction

[0402] Predictive information can be used to generate a predicted block for a target block.

[0403] The encoding device (110) can generate prediction information required for prediction and can generate a bitstream containing the prediction information. The prediction information can be signaled from the encoding device (110) to the decoding device (150) through the bitstream. The decoding device (150) can obtain the prediction information from the bitstream and can generate a prediction block by performing a prediction on a target block using the prediction information.

[0404] Prediction information may include intra prediction information, inter prediction information, and IBC prediction information. In the embodiments, prediction information may be replaced with intra prediction information, inter prediction information, and / or IBC information. Intra prediction information may include information used for intra prediction as described in the embodiments. Inter prediction information may include information used for inter prediction as described in the embodiments. IBC information may include information used for IBC prediction as described in the embodiments.

[0405]

[0406] Intra prediction

[0407] Figure 3 shows the structure of an intra prediction according to one embodiment.

[0408] Intra-prediction can be performed using reference samples and coding parameters of the target block. The reference sample may be a (restored) sample within the (restored) reference block. Alternatively, an intermediate prediction sample may be generated using a sample described in an embodiment, such as the restored sample, and a reference sample may be generated again using the intermediate prediction sample. Processing described in an embodiment, such as filtering, may be applied when generating the reference sample.

[0409] The reference block may be a (spatial) neighbor block of the target block. The coding parameter may be a coding parameter for the target block and / or a coding parameter for the reference block. In intra-prediction, the reference sample may refer to a neighbor sample.

[0410] A prediction block can be generated by performing intra prediction on a target block according to an intra prediction mode, based on a reference sample within the target image and information related to the reference sample. The size of the target block and the size of the prediction block may be the same.

[0411] In the embodiments, the prediction block may be a PU. Alternatively, the prediction block may correspond to the CU or TU described in the embodiments. The prediction block may have a square or rectangular shape.

[0412] An intra prediction mode can be represented by at least one of a mode number, a mode value, a mode angle, and a mode direction. The prediction directions of a plurality of intra prediction modes for a target block are illustrated in the lower right corner of FIG. 3. Among the plurality of intra prediction modes, the remaining intra prediction modes, excluding DC and planar modes, may be directional modes. A directional mode may be an intra prediction mode having a specific direction or a specific angle. An intra prediction mode for a target block may be selected from directional modes and non-directional modes.

[0413] In the bottom-right rectangle representing the target block, the number '0' may represent Planner mode, a non-directional intra prediction mode. The number '1' may represent DC mode, a non-directional intra prediction mode. In the bottom-right rectangle representing the target block, arrows extending from the center of the rectangle outwards may represent the prediction directions of directional intra prediction modes. Additionally, a number displayed near an arrow may represent an example of a mode value assigned to an intra prediction mode or the prediction direction of an intra prediction mode.

[0414] Intra prediction can be performed according to the intra prediction mode for the target block. One of the intra prediction modes available for the target block can be used as the intra prediction mode for the target block.

[0415] The number of intra prediction modes available to the target block may be a predefined value. Alternatively, the number of intra prediction modes available to the target block may be determined based on the attributes of the prediction block. For example, the attributes of the prediction block may include coding parameters such as shape, size, and color components.

[0416] For example, in FIG. 3, the directional modes illustrated by dashed lines (i.e., directional modes with numbers from -14 to -1 or numbers from 67 to 80) can be applied only to predictions for non-square blocks. Therefore, the number of available intra-prediction modes for predictions for square blocks may be 67. (Planner mode, DC mode, and 65 directional modes)

[0417] For example, the number of available intra prediction modes may vary depending on whether the color component of a block is a luminance signal or a chroma signal. The number of available intra prediction modes for a block with a luminance component may be greater than the number of available intra prediction modes for a block with a chroma component.

[0418] Intra-prediction modes may include a horizontal-below mode, a horizontal mode, a vertical mode, and a vertical-right mode. The horizontal-below mode may be an intra-prediction mode located at the bottom of the horizontal mode. The vertical-right mode may be a mode located to the right of the vertical mode. For example, in FIG. 3, the mode value of the horizontal mode may be 18. The mode value of the vertical mode may be 50. Intra-prediction modes with a mode value of 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, and 66 may be vertical-right modes. Intra prediction modes with a mode value of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 and 17 may be horizontal bottom modes.

[0419] The number of the aforementioned intra-prediction modes and the mode number of each intra-prediction mode may be exemplary only. The number of the aforementioned intra-prediction modes and the mode number of each intra-prediction mode may be defined differently depending on the embodiment, implementation, and / or as necessary.

[0420] When the intra prediction mode is a planner mode, when generating a prediction block of a target block, the sample value of the prediction sample can be generated using a weighted sum (weighted sum) of the top reference sample of the target sample, the left reference sample of the target sample, the right top reference sample of the target block, and the left bottom reference sample of the target block, depending on the position of the prediction sample within the prediction block.

[0421] When the intra prediction mode is DC mode, a prediction block can be generated based on the average of the sample values ​​of multiple reference samples. The multiple reference samples may include top reference samples and left reference samples of the target block. The value of the prediction sample of the prediction block can be determined based on the average of the sample values ​​of the multiple reference samples. Additionally, filtering using the values ​​of the reference samples can be performed on specific rows and / or specific columns within the target block. The specific rows may be one or more top rows adjacent to the top reference samples. The specific columns may be one or more left columns adjacent to the left reference samples.

[0422] When the intra prediction mode is a directional mode, a prediction block can be generated using the top reference sample, left reference sample, right top reference sample, and / or left bottom reference sample of the target block.

[0423] The intra prediction mode of a target block can be determined based on the intra prediction mode of a neighboring block of the target block. Information for determining the intra prediction mode of the target block can be signaled.

[0424] For example, if the intra prediction modes of the target block and the neighbor block are the same, an indicator indicating that the intra prediction modes of the target block and the neighbor block are the same may be signaled.

[0425] For example, an indicator indicating an intra prediction mode such as the intra prediction mode of the target block among the intra prediction modes of multiple neighboring blocks may be signaled.

[0426] For example, if the intra prediction modes of the target block and neighboring blocks are different from each other, an indicator indicating the intra prediction mode of the target block may be signaled. Alternatively, information used to derive the intra prediction mode of the target block based on the intra prediction mode of the neighboring block may be signaled.

[0427] Reference samples used for intra-prediction of a target block may include bottom-left reference samples, left reference samples, top-left reference samples, top reference samples, and top-right reference samples, etc.

[0428] For example, left reference samples may be restored reference samples adjacent to the left side of the target block. Top reference samples may be restored reference samples adjacent to the top side of the target block. Top-left reference samples may be restored reference samples diagonally adjacent to the top-left side of the target block. Bottom-left reference samples may be reference samples located below the left reference samples among samples located on the same line as the left sample line composed of left reference samples. Top-right reference samples may be reference samples located to the right of the top reference samples among samples located on the same line as the top sample line composed of top reference samples.

[0429] Reference samples used for intra-prediction for a target block can be determined based on the intra-prediction mode of the target block. One or more reference samples may be used to determine the sample values ​​of the prediction samples of the prediction block. In FIG. 3, the direction of the intra-prediction mode indicated by the arrow may represent the direction from the prediction sample to the reference sample. The direction of the intra-prediction mode may represent the dependency relationship between the reference samples and the prediction samples. For example, depending on the intra-prediction mode, the sample value of a specific reference sample may be used as the sample value of at least one sample of the prediction block. Here, the specific reference sample and at least one sample of the prediction block may be samples designated by a straight line of the direction of the intra-prediction mode. That is to say, the sample value of the specific reference sample may be copied to the sample value of the prediction sample located in the reverse direction of the direction of the intra-prediction mode. Alternatively, the sample value of the prediction sample of the prediction block may be the sample value of the reference sample located in the direction of the intra-prediction mode relative to the location of the prediction sample.

[0430] Reference samples used for intra-prediction may not be limited to samples immediately adjacent to the target block. As illustrated in FIG. 3, at least one of reference sample line 0 to reference sample line 3 may be used for intra-prediction of the target block.

[0431] Each reference sample line in FIG. 3 may contain one or more reference samples. The smaller the number of the reference sample line, the closer the line of reference samples may be to the target block. Reference sample line 0 may be a line of reference samples immediately adjacent to the target block. When the top-left coordinates of the target block are (X, Y), the horizontal length is W, and the vertical length is H, the reference samples of reference sample line 0 may be samples with an x-coordinate of X-1 or a y-coordinate of Y-1. Here, the y-coordinates of the reference samples with an x-coordinate of X-1 may be Y-1 to Y+2H. The x-coordinates of the reference samples with a y-coordinate of Y-1 may be X-1 to X+2W. The reference samples of reference sample line A may be samples with an x-coordinate of XA-1 or a y-coordinate of YA-1. Here, the y-coordinates of the reference samples with an x-coordinate of XA-1 may be YA-1 to Y+2H+A. The x-coordinates of reference samples with y-coordinate YA-1 can be XA-1 to X+2W+A. A can be 1, 2, or 3.

[0432] Samples of segments A and F can be derived using padding that uses the nearest samples of segments B and E, respectively, instead of being obtained from restored neighbor blocks.

[0433] The reference sample line index may indicate a reference sample line among multiple reference sample lines used for intra-prediction of a target block. For example, the reference sample line index may have a value from 0 to 3. The reference sample line index may be signaled.

[0434] When intra-color component prediction is used for a target block, a prediction block for a second color component can be generated based on a reconstruction block of a first color component for the target block. For example, the first color component may be a luminance component, and the second color component may be a chroma component.

[0435] For intra-prediction between color components, parameters between the first and second color components can be derived based on a template. For example, the parameters can be parameters of a linear model.

[0436] For example, the template may include a top reference sample and / or a left reference sample of the target block, and may include a top reference sample and / or a left reference sample of the restoration block of the first color component corresponding to these reference samples.

[0437] Once the parameters are derived, a prediction block for a second color component for a target block can be generated by applying the reconstruction block of the first color component to a linear model. Depending on the image format or the type of intra-prediction between color components, subsampling or downsampling may be performed on the surrounding samples of the reconstruction block of the first color component and on the reconstruction block of the first color component. If subsampling is performed, the derivation of parameters and the intra-prediction between color components may be performed using corresponding samples derived by subsampling.

[0438] Intra Sub-Partitions (ISP) prediction may refer to sequential intra prediction for multiple subblocks generated by partitioning a target block. In ISP prediction, the target block may be partitioned into two or four subblocks in the horizontal and / or vertical directions. The partitioned subblocks may be restored sequentially. As intra prediction is performed on the subblocks, sub-prediction blocks for the subblocks may be generated. Additionally, as inverse quantization and / or inverse transformation is performed on the subblocks, sub-residual blocks for the subblocks may be generated. A restored subblock may be generated by adding the sub-prediction blocks to the sub-residual blocks. The restored subblocks may be used as reference samples for intra predictions for other subblocks to be processed next.

[0439] In performing a prediction for a target block, it may be determined whether samples included in a restored neighbor block can be used as reference samples for the target block. If there are non-available samples among the samples in the neighbor block that cannot be used as reference samples for the target block, a value generated by copying and / or interpolation using the sample value of at least one sample among the samples included in the restored neighbor block may replace the sample value of the non-available sample. If the value generated by copying and / or interpolation replaces the sample value of the sample, the sample may be used as a reference sample for the target block.

[0440] In intra-prediction, the sample value of a prediction sample in a prediction block can be determined by the sample value of a reference sample. The location of the reference sample can be specified by the location of the prediction sample and the direction of the intra-prediction mode. If the location specified by the location of the prediction sample and the direction of the intra-prediction mode is an integer location, the sample value of one reference sample pointed to by the integer location can be used to determine the sample value of the prediction sample in the prediction block. If the location specified by the location of the prediction sample and the direction of the intra-prediction mode is not an integer location, an interpolated reference sample can be generated based on the two reference samples closest to the specified location. The sample value of the interpolated reference sample can be used to determine the sample value of the prediction sample. That is to say, when the location specified by the location of the prediction sample and the direction of the intra-prediction mode represents the space between two reference samples, an interpolated sample value can be generated based on the sample values ​​of the two samples.

[0441]

[0442] Inter prediction

[0443] FIG. 4 shows the structure of an inter prediction to explain an inter prediction process according to one embodiment.

[0444] The rectangle shown in Fig. 4 can represent an image. Additionally, the arrow in Fig. 4 can indicate the predicted direction.

[0445] Each image constituting a video can be classified into I-pictures (i.e., intra-pictures), P-pictures (i.e., uni-prediction pictures), and B-pictures (i.e., bi-prediction pictures) according to their coding type. Coding can be performed for each picture according to its coding type.

[0446] If the target picture is an I picture, coding for the target picture can be performed using information within the target picture without inter-prediction referencing other images. For example, coding for the I picture can be performed using intra-prediction and / or IBC prediction.

[0447] Coding for P picture and B picture can be performed by at least one of intra prediction, IBC prediction, and inter prediction using a reference image.

[0448] If the target picture is a P picture, coding for the target picture can be performed using unidirectional inter-prediction using a single reference image list.

[0449] When the target picture is picture B, coding for the target picture can be performed using unidirectional inter-prediction or bidirectional inter-prediction using two reference image lists.

[0450] Below, the inter prediction for the target block in the inter mode according to the embodiment is described in detail.

[0451] When the prediction mode of the target block is inter mode, inter prediction can be performed on the target block. The target block can be a prediction block or a partitioned prediction block.

[0452] Inter prediction can be performed using reference images and motion information. In inter prediction, a reference image can be selected using a reference image index, and a reference block corresponding to a target block within the reference image can be determined using motion information. A prediction block for the target block can be generated using the determined reference block.

[0453] Motion information can be derived using coding parameters, etc. For example, motion information can be derived using motion information of restored neighbor blocks, motion information of call blocks, and / or motion information of blocks adjacent to call blocks.

[0454] In the embodiments, a candidate list may be used for inter prediction. The candidate list may include multiple candidates. An index pointing to a candidate among the candidates in the candidate list that is used for inter prediction for a target block may be signaled. The candidate list may be derived in the same manner based on the same information in the encoding device (110) and the decoding device (150). Here, the same information may include a restored image and a restored block. Additionally, in order to specify a candidate by an index, the order of candidates within the candidate list may be constant.

[0455] In one embodiment, a prediction for a target block can be performed by using motion information of a spatial candidate or a temporal candidate as motion information of the target block. Motion information of a spatial candidate may be referred to as spatial motion information. Motion information of a temporal candidate may be referred to as temporal motion information.

[0456] Spatial candidates may be restored spatial neighbor blocks that are spatially adjacent to the target block.

[0457] Spatial candidates may be blocks that 1) exist within the target image, 2) have already been restored through decoding, and 3) are adjacent to the target block.

[0458] Spatial candidates may include the left block, top block, bottom-left block, top-right block, and top-left block of the target block.

[0459] Temporal candidates may be restored temporal neighbor blocks corresponding to the target block within the restored call (COL) image.

[0460] In the embodiments, the motion information of the spatial candidate may be the motion information of a block containing the spatial candidate. The motion information of the temporal candidate may be the motion information of a block containing the temporal candidate.

[0461] In inter prediction, a call block for a target block can be identified. The region of the target block within the target image and the region of the call block within the call image may be the same. That is to say, the call block may be a block that occupies a specific region within the call image. The specific region may be a region corresponding to the region of the target block within the call image.

[0462] Temporal candidates may be locations inside and / or outside the call block within the call image.

[0463] For example, a call block may include a first call block and a second call block. When the top-left coordinates of the call block are (xP, yP) and the size of the call block is (nPSW, nPSH), the first call block may be a block occupying the coordinates (xP + nPSW, yP + nPSH). The second call block may be a block occupying the coordinates (xP + (nPSW >> 1), yP + (nPSH >> 1)). The second call block may optionally be used as a call block when the first call block is unavailable.

[0464] The MV of the target block can be determined based on the MV of the call block. Scaling can be performed on the MV of the call block. The scaled MV of the call block can be used as the MV of the target block or the prediction MV. Alternatively, the MV of the temporal candidate stored in the candidate list associated with the inter-prediction can be the scaled MV.

[0465] The ratio of the scaled MV and the MV of the call block may be equal to the ratio of the first temporal distance and the second temporal distance. The first temporal distance may be the distance between the reference image of the target block and the target image. The second temporal distance may be the distance between the reference image of the call block and the call image.

[0466] The method by which motion information is derived can be determined by the inter prediction mode of the target block. For example, as an inter prediction mode, AMVP mode, merge mode, skip mode, merge mode with MVD, sub-block merge mode, GPM, Combined Inter Intra Prediction (CIIP) mode, and affine inter mode may be used. In the following embodiments, each of the inter prediction modes is described.

[0467]

[0468] AMVP mode

[0469] When AMVP mode is used as a prediction mode, a list of MV candidates including one or more MV candidates can be generated using spatial candidate MVs, temporal candidate MVs, history-based MV candidates, and zero vectors. At least one of the spatial candidate MVs, temporal candidate MVs, and zero vectors can be determined and used as an MV candidate.

[0470] Spatial candidates may include restored spatial neighbor blocks. The MV of a restored spatial neighbor block may be referred to as a spatial motion vector candidate. Temporal candidates may include a call block and a block adjacent to the call block. The MV of a call block or the MV of a block adjacent to the call block may be referred to as a temporal motion vector candidate. History-based MV candidates may be MVs in a list containing MVs of other blocks that were encoded / decoded before the encoding / decoding of the target block.

[0471] The encoding device (110) can determine the MV to be used for encoding a target block within a search range using an MV candidate list. The maximum number of MV candidates in the MV candidate list may be predefined. N may represent the predefined maximum number. For example, N may be 2. Alternatively, the maximum number of such candidates may be signaled from the encoding device to the decoding device or derived from the decoding device. The encoding device (110) can determine an MV candidate to be used as the predicted MV of the target block among the MV candidates in the MV candidate list. The MV to be used for encoding the target block may be an MV that can be encoded at the minimum cost. The encoding device (110) may determine whether to use the AMVP mode in encoding the target block and may generate AMVP mode usage information indicating whether the AMVP mode is used.

[0472] Inter prediction information may include 1) AMVP mode usage information, 2) MV candidate index, 3) MVD, 4) MVD resolution information, 5) reference direction and 6) reference image index, and may include residual blocks. Inter prediction information may be signaled from the encoding device (110) to the decoding device (150) in the form of a bitstream.

[0473] The decoding device (150) can obtain AMVP mode usage information from the bitstream. If the AMVP mode usage information indicates that the AMVP mode is being used, the decoding device (150) can obtain an MV candidate index, an MVD, MVD resolution information, a reference direction, and a reference image index from the bitstream. Among the MV candidates included in the MV candidate list, the MV candidate pointed to by the MV candidate index can be selected as the predicted MV of the target block.

[0474] The MVD may represent the difference between the MV that will actually be used for inter-prediction of the target block and the predicted MV. The encoding device (110) may derive a predicted MV that is close to the MV that will actually be used for inter-prediction of the target block in order to use an MVD of the smallest possible size. The decoding device (150) may derive the MV of the target block by summing the MVD and the predicted MV. That is to say, the MV of the target block derived by the decoding device (150) may be the sum of the MVD and the predicted MV candidates.

[0475] Additionally, the encoding device (110) can generate MVD resolution information. The MVD resolution information may be information used to adjust the resolution of the MVD. The decoding device (150) can adjust the resolution of the MVD using the MVD resolution information.

[0476] Meanwhile, the encoding device (110) can calculate the MVD based on an affine model. The affine control point MV of the target block can be derived based on the sum of the affine control point MV candidates and the MVD. Using the affine control point MV, the MV of each sub-block within the target block can be derived.

[0477]

[0478] Merge Mode

[0479] When merge mode is used, a merge candidate list containing multiple merge candidates can be generated using motion information of spatial candidates and motion information of temporal candidates, etc. Motion information may include 1) MV, 2) reference image index and 3) reference direction, etc. A merge candidate may be motion information.

[0480] Merge candidates may include 1) spatial merge candidates generated based on spatial candidates, 2) temporal merge candidates generated based on temporal candidates, 3) history-based merge candidates, 4) average merge candidates, and 5) zero merge candidates.

[0481] A history-based merge candidate may be movement information within a list containing movement information of other blocks that were encoded / decoded earlier than the encoding / decoding of the target block.

[0482] The average merge candidate may be a merge candidate generated based on the average of two merge candidates within the merge candidate list.

[0483] Zero merge candidates can be zero vector motion information. Zero vector motion information can be motion information where MV is a zero vector.

[0484] Merge candidates can be added to the merge candidate list according to a predefined method and a predefined order so that the merge candidate list has a set number of merge candidates. The same merge candidate list can be configured in the encoding device (110) and the decoding device (150) through the predefined method and a predefined order.

[0485] The encoding device (110) can select a merge candidate to be used for encoding a target block from among the merge candidates in the merge candidate list. The encoding device (110) can determine whether to use a merge mode in encoding the target block and can generate merge mode usage information indicating whether the merge mode is used.

[0486] Inter prediction information may include 1) merge mode usage information, 2) merge index and 3) correction information, etc., and may include residual blocks. Inter prediction information may be signaled from the encoding device (110) to the decoding device (150) in the form of a bitstream.

[0487] The decoding device (150) can obtain merge mode usage information from the bitstream. If the merge mode usage information indicates that the merge mode is being used, the decoding device (150) can obtain merge mode-related information, such as a merge index, from the bitstream.

[0488] The encoding device (110) can select the optimal merge candidate among the merge candidates included in the merge candidate list and can set the value of the merge index to point to the selected merge candidate.

[0489] Correction information may be information used for correcting the MV. The encoding device (110) may generate correction information. The decoding device (150) may derive a corrected MV by performing correction on the MV of a merge candidate selected by a merge index based on the correction information. The corrected MV may be used as the MV of the target block.

[0490] In one embodiment, the correction information may include an MVD. The correction information may include one or more of correction usage information, correction direction information, and correction magnitude information. The correction usage information may indicate whether to use correction for the MV. A merge mode that performs correction for the MV based on the correction information may be referred to as a merge mode having an MVD.

[0491] In merge mode, a prediction for the target block can be performed using the merge candidate pointed to by the merge index among the merge candidates included in the merge candidate list.

[0492] Movement information of the target block can be derived from 1) MV, 2) reference image index and 3) reference direction of the merge candidate pointed to by the merge index.

[0493] In one embodiment, the merge candidates in the merge candidate list may be specific modes that induce inter-prediction information. A merge candidate may be information pointing to a specific mode that induces inter-prediction information. Inter-prediction information of a target block may be induced according to the specific mode pointed to by the merge candidate. In this regard, the specific mode may be regarded as a specific inter-prediction information inducing mode or a specific movement information inducing mode. The specific mode may include a series of processes that induce inter-prediction information.

[0494] Inter-prediction information of the target block can be derived according to a specific mode pointed to by a merge candidate selected by a merge index among the merge candidates in the merge candidate list. For example, specific modes may include a mode for deriving motion information at the sub-block level and a mode for deriving motion information at the affine level, and may include other modes for deriving motion information as described in the embodiments.

[0495] Skip mode may be a mode that does not use residual blocks. That is to say, when skip mode is used, the restoration block may be identical to the prediction block. The description of the merge mode in the embodiments may also apply to skip mode. The difference between merge mode and skip mode may be whether or not residual blocks are signaled and used. That is to say, skip mode may be similar to merge mode except that residual blocks are not transmitted / used, and the description of merge mode may also apply to skip mode.

[0496] The subblock merge mode may be a mode in which motion information of a target subblock is induced for a target subblock within a target block. When the subblock merge mode is applied, a list of subblock merge candidates may be generated using affine control point motion vector merge candidates and / or subblock-based temporal merge candidates. The subblock-based temporal merge candidates may be motion information of the call subblock of the target subblock.

[0497] In GPM, a first prediction block and a second prediction block can be generated using two sets of motion information for a target block. For each coordinate of the target block, a final prediction sample of the final prediction block can be generated using the weighted sum of the first prediction sample of the first prediction block and the second prediction sample of the second prediction block.

[0498] Here, the first weight for the first prediction sample of the weighted consensus and the second weight for the second prediction sample can be determined based on the boundaries of the GPM. The boundaries may represent dividing lines that divide the target block. Depending on the boundaries, the target block may be divided into a first divided region and a second divided region.

[0499] If the distance between the final prediction sample and the boundary is less than or equal to a reference value, the value of the final prediction sample of the final prediction block may be determined using the weighted sum of the first prediction sample of the first prediction block and the second prediction sample of the second prediction block. If the distance between the final prediction sample and the boundary is greater than the reference value, one of the first weight and the second weight may be 1 and the other may be 0.

[0500] The Combined Inter-Intra Prediction (CIIP) mode may be a mode that derives a prediction sample of a target block using a weighted sum of a prediction sample generated by inter-prediction and a prediction sample generated by intra-prediction.

[0501] In the aforementioned modes, self-improvement of the derived motion information may be performed, and the improved motion information may be used as motion information for the target block. For example, blocks within a specific region determined based on the derived motion information may be searched, and the motion information of the block having the smallest Sum of Absolute Differences (SAD) value among the searched blocks may be used as the improved motion information for the target block. The specific region may be a square region within the reference image specified by the motion information. The point indicated by the motion information may be the center of the specific region.

[0502] In the aforementioned modes, compensation for prediction samples derived through inter-prediction can be performed using optical flow.

[0503]

[0504] FIG. 5 shows the order of addition of spatial candidates to the candidate list according to one embodiment.

[0505] In Fig. 5, the locations of the spatial candidates are shown.

[0506] The large block in the center can represent the target block. The five small blocks adjacent to the target block can represent spatial candidates.

[0507] The coordinates of the target block can be (xP, yP), and the size of the target block can be (nPSW, nPSH).

[0508] Spatial candidate A0 may be a block adjacent to the bottom-left of the target block. A0 may be a block occupying a sample of coordinates (xP - 1, yP + nPSH).

[0509] Spatial candidate A1 may be a block adjacent to the left of the target block. A1 may be the bottommost block among the blocks adjacent to the left of the target block. Or, A1 may be a block adjacent to the top of A0. A1 may be a block occupying a sample of coordinates (xP - 1, yP + nPSH - 1).

[0510] Spatial candidate B0 may be a block adjacent to the top right of the target block. B0 may be a block occupying a sample of coordinates (xP + nPSW, yP - 1).

[0511] Spatial candidate B1 may be a block adjacent to the top of the target block. B1 may be the rightmost block among the blocks adjacent to the top of the target block. Or, B1 may be a block adjacent to the left of B0. B1 may be a block occupying a sample of coordinates (xP + nPSW - 1, yP - 1).

[0512] Spatial candidate B2 may be a block adjacent to the top-left corner of the target block. B2 may be a block occupying a sample of coordinates (xP - 1, yP - 1).

[0513] As illustrated in Fig. 5, in adding spatial candidates to the candidate list, B1, A1, The order B0, A0, and B2 can be used. That is, B1, A1, Available spatial candidates can be added to the candidate list in the order of B0, A0, and B2. The order in which spatial candidates illustrated in FIG. 5 are added to the merge candidate list may be just one example.

[0514] The above candidate list may include a motion information candidate list, a merge candidate list, an MV candidate list, a BV candidate list, and an MPM list, etc.

[0515] To include a spatial or temporal candidate in the candidate list, it may be determined whether the spatial or temporal candidate is available. If a candidate block is outside the boundaries of an image, slice, or tile, the availability of the candidate block may be set to false. The description "availability is set to false" may mean "it is set to non-available."

[0516] The maximum number of candidates in the candidate list can be set. N can represent the set maximum number. The set maximum number can be signaled through a parameter set or header, etc. For example, the maximum number of candidates in the candidate list for a target block within a slice can be set by the slice header. For example, the value of N can be 5 by default.

[0517]

[0518] IBC mode

[0519] The IBC mode may be an intra-block copy prediction mode that generates a prediction block for a target block by referencing an already reconstructed region within the target image. In this respect, the IBC mode may also be referred to as a current image reference mode. A block vector (BV) may be used to identify the already reconstructed region.

[0520] Whether the target block is encoded / decoded in IBC mode can be determined using IBC mode usage information. The encoding device (110) can determine whether to use IBC mode in encoding the target block and can generate IBC mode usage information indicating whether IBC mode is used. The decoding device (150) can obtain IBC mode usage information from the bitstream.

[0521] In IBC mode, the predicted block of the target block can be generated based on the BV. The BV can specify the reference block. The BV can indicate the displacement between the target block and the reference block. The reference block can be a block within the target image. The description of the MV of the embodiments can also be applied to the BV.

[0522] The IBC mode may include a skip mode, a merge mode, and an AMVP mode, etc. The descriptions of the AMVP mode, merge mode, and skip mode of the embodiments may be similarly applied to the AMVP mode, merge mode, and skip mode of the IBC mode, respectively.

[0523] In skip mode or merge mode, a merge candidate list may be configured, and a merge index may specify one merge candidate from among the merge candidates in the merge candidate list. The BV of the specified merge candidate may be used as the BV of the target block.

[0524] In AMVP mode, BVD can be used. The description of MVD in the embodiments can also be applied to BVD.

[0525] The reference block in IBC mode may be limited to a block within an already restored region of the target image. Alternatively, the reference block may be contained within at least one of the target CTU or the left CTUs. For example, the value of BV may be restricted so that the reference block is located within a specific region. The specific region may be an area of ​​three blocks of a specific size that are encoded / decoded before the block of a specific size containing the target block. The specific size may be 64x64.

[0526]

[0527] Transformation and Quantization

[0528] Quantized levels can be generated by performing a transformation and / or quantization on the residual block. The residual block can represent the difference between the original block and the prediction block. A restored residual block can be generated by performing inverse quantization and / or inverse transformation on the quantized levels. The restored residual block can represent the difference between the restored block and the prediction block.

[0529] When a transformation or inverse transformation is performed, a separable transform or a 2D non-separable transform may be performed on the residual block. A separable transform may be a transformation that performs 1D transformations on the residual block in the horizontal and vertical directions, respectively.

[0530] The transformation kernels used for the transformation may include various DCT kernels such as DCT type 2 (DCT-II), 2) DST kernels, and 3) kernels derived by training. For 1D transformation, DCT type and DST type may include DCT-V, DCT-VIII, DST-I, and DST-VII in addition to DCT-II.

[0531] A transformation set may be used to determine the DCT type, DST type, or learning-derived kernel to be used for the transformation. Each transformation set may include multiple transformation candidates. Each transformation candidate may be a DCT type, a DST type, or a learning-derived kernel, etc.

[0532] The encoding device (110) can perform conversion and inverse conversion using conversion candidates included in the conversion set. The decoding device (150) can perform inverse conversion using conversion candidates included in the conversion set. Conversion selection information indicating which conversion candidate is used among the plurality of conversion candidates included in the conversion set applied to the residual block may be signaled. The conversion selection information may include vertical conversion selection information and horizontal conversion selection information. The vertical conversion selection information may indicate which conversion among the conversions belonging to the conversion set is used for the vertical conversion. The horizontal conversion selection information may indicate which conversion among the conversions belonging to the conversion set is used for the horizontal conversion.

[0533] The transformation may include at least one of a primary transformation and a secondary transformation. A primary transformation coefficient may be generated by performing a primary transformation on a residual block, and a secondary transformation coefficient may be generated by performing a secondary transformation on the transformation coefficient. Here, the transformation coefficient may include a primary transformation coefficient and a secondary transformation coefficient.

[0534] A first-order transformation may mean a Multiple Transform Selection (MTS) that applies different transformations to each of the 1D directions (i.e., vertical and horizontal directions).

[0535] A second-order transformation may be a transformation intended to improve the energy concentration of the transformation factors generated by a first-order transformation. A second-order transformation may be 1) a separable transformation like the first-order transformation, or 2) a 2D non-separable transformation. A 2D non-separable transformation may refer to a Low Frequency Non-Separable Transform (LFNST) or a Non-Separable Primary Transform (NSPT).

[0536] NSPT can be applied to specific block sizes such as 4x4, 4x8, 8x4, 4x16, 16x4, 8x8, 8x16, and 16x8 for intra-coding.

[0537] A first-order transformation may be performed using at least one of a plurality of predefined transformation methods. For example, the plurality of predefined transformation methods may include DCT, DST, and KLT, etc. Additionally, the first-order transformation may be a transformation having various transformation types according to transformation kernel functions that define DCT and DST. For example, the first-order transformation may include a plurality of transformations such as DCT-2, DCT-4, DCT-5, DCT-7, DCT-8, DST-1, DST-2, DST-4, DST-7, and DST-8 according to a plurality of transformation kernels.

[0538] In one embodiment, the transformation type may be determined based on coding parameters related to the target block. For example, the transformation type may be determined based on one or more of 1) the prediction mode of the target block (e.g., one of intra prediction and inter prediction), 2) the size of the target block, 3) the shape of the target block, 4) the intra prediction mode of the target block, 5) the components of the target block (e.g., one of luminance components and chroma components), and 6) the splitting type applied to the target block (e.g., one of QT, BT, TT, and non-split).

[0539] As in the first transformation, a set of transformations can be defined in the second transformation as well. Methods for deriving and / or determining the set of transformations of the embodiments can be applied to the second transformation as well as the first transformation.

[0540] In one embodiment, the first transformation and / or second transformation may be determined for a specific target. The transformation selection information may include transformation target information. The transformation target information may indicate the target to which the first transformation and / or second transformation is applied.

[0541] For example, first-order transformation and / or second-order transformation may be applied to one or more signal components among the luminance component and the chroma component.

[0542] In one embodiment, the transformation selection information may include first transformation usage information and second transformation usage information. The first transformation usage information may indicate whether a first transformation is applied to the residual block of the target block. The second transformation usage information may indicate whether a second transformation is applied to the residual block of the target block.

[0543] In one embodiment, whether a first transformation and / or a second transformation is applied may be determined based on coding parameters for the target / neighbor block, such as the size and shape of the target / neighbor block.

[0544] In one embodiment, the transformation selection information may include first transformation selection information and second transformation selection information. The first transformation selection information may indicate a transformation method applied to a residual block among a plurality of transformation methods that can be used in the first transformation. The first transformation selection information may be a first transformation index. The second transformation selection information may indicate a transformation method applied to a transformation coefficient among a plurality of transformation methods that can be used in the second transformation. The second transformation selection information may be a second transformation index.

[0545] In one embodiment, the transformation methods of the first transformation and the second transformation can each be derived based on specific information such as coding parameters. For example, the coding parameters may include coding parameters for target / neighbor blocks.

[0546] In the embodiments, information related to a transformation, such as transformation selection information, and sub-information of the transformation selection information may be signaled to a specific target. For example, the specific target may be a CU.

[0547] Information related to transformations, such as transformation selection information, and sub-information of transformation selection information can be derived for a specific target. For example, the specific target may be a CU.

[0548] Quantized levels can be generated by performing quantization on the result or residual block generated by performing a first-order transformation and / or a second-order transformation.

[0549] The description of the transformation described above may also be applied to the inverse transformation. In such application, the inverse processing of the processing described for the transformation may be performed in the inverse transformation. "Transformation" within the name related to the transformation may be changed to "inverse transformation." Additionally, the input of the transformation may be considered as the output of the inverse transformation. The output of the transformation may be considered as the input of the inverse transformation. The decoding device (150) may obtain information related to the transformation, such as transformation selection information, and may use the information related to the transformation to perform the inverse processing of the processing related to the transformation indicated by the information related to the transformation.

[0550] The target block may include multiple subblocks. Each subblock may be defined according to a minimum block size or minimum block shape. The target block may be divided into multiple subblocks, and each subblock may include coefficients such as 4x4, 2x8, and 8x2. The target block may be a transformation block. Transform coefficients or quantized levels may be represented in the form of a block. Transform coefficients may be quantized transformation coefficients.

[0551] Transform coefficients or quantized levels may be scanned according to at least one scanning type among diagonal scanning, vertical scanning, and horizontal scanning. Diagonal scanning may be top-right diagonal scanning or bottom-left diagonal scanning.

[0552] For example, by scanning the coefficients of a block using diagonal scanning, the coefficients can be changed or arranged into a one-dimensional vector form. Vertical scanning may be scanning the coefficients in the form of a two-dimensional block in a column direction. Horizontal scanning may be scanning the coefficients in the form of a two-dimensional block in a row direction.

[0553] The scanning type for the coefficients can be determined based on coding parameters such as intra prediction mode, block size, and block shape. For example, based on coding parameters such as intra prediction mode, block size, and block shape, it can be determined which scanning method—diagonal scanning, vertical scanning, and horizontal scanning—will be used. A block may be a transformation unit.

[0554] Scanning according to each scanning type can start at a specific starting point and end at a specific ending point.

[0555] In scanning, the scanning order according to the scanning type can first be applied between subblocks. Next, the scanning order according to the scanning type can be applied to the transformation coefficients or quantized levels within the subblocks.

[0556] The encoding device (110) can perform entropy encoding on the conversion coefficients or quantized levels to generate a bitstream containing entropy-encoded conversion coefficients or entropy-encoded quantized levels.

[0557] The decoding device (150) can generate the transform coefficients or quantized levels by obtaining entropy-encoded transform coefficients or entropy-encoded quantized levels from the bitstream and performing entropy decoding. The coefficients can be arranged in the form of two-dimensional blocks through inverse scanning. The arrangement of inverse scanning may be a rearrangement opposite to the arrangement of scanning.

[0558] Backscanned transform coefficients or backscanned quantized levels can be generated through backscanning of the coefficients. In this case, the backscanning types of backscanning may include diagonal scans, vertical scans, and horizontal scans, and a backscanning type of the inverse transform corresponding to the scanning type of the transform may be selected.

[0559] In the decoding device (150), inverse quantization can be performed on the (backscanned) coefficients. Depending on whether a second inverse transform is performed, a second inverse transform can be performed on the result generated by the performance of inverse quantization. Also, depending on whether a first inverse transform is performed, a first inverse transform can be performed on the result generated by the performance of the second inverse transform. By selectively performing a second inverse transform and a first inverse transform on the coefficients, a restored residual block can be generated.

[0560]

[0561] Filtering

[0562] To improve the image quality, filtering may be performed on the blocks. The value of the target sample may be determined or updated by the filtering.

[0563] The target sample may be one of the samples described in the embodiments. For example, the target sample may be one or more of the samples described in the embodiments, such as a prediction sample, a reference sample, a residual sample, a reconstructed sample, and a reconstructed sample to which filtering has been applied.

[0564] The target sample may be a sample within one or more of the target picture, target slice, target CTB, target block, reference sample line, and template. The target block may be one of the blocks described in the embodiments. For example, the target block may be one or more of the blocks described in the embodiments, such as a transformation block, prediction block, reference block, residual block, and restoration block.

[0565] In the embodiments, the filtering process described as being applied to one target may also be applied to other targets. For example, the filtering process described in a specific in-loop filtering may also be applied to a transformation block, a prediction block, a reference block, and a residual block, etc.

[0566] For the filtering of the embodiments, a specific type of filtering may be used. The type of filtering may include a filter tap (or filter tap length), a filter shape, a filter strength, filter coefficients (or weights), and an offset.

[0567] The filter tab may indicate the number of input samples used for the filter. The input samples may include target samples. Alternatively, the input samples may include specific values ​​determined for the target samples. The input samples may include one or more reference samples. One or more reference samples may be determined based on the attributes of the target block described in the embodiments. The attributes may include coding parameters. For example, the attributes of the target sample may include the location of the target sample. One or more reference samples may be specified based on their relative position to the location of the target sample.

[0568] The filter shape can represent the shape formed by input samples. A specific value determined for a target sample can be considered as the target sample. In other words, if a specific value determined for a target sample is used as an input sample for a filter, the target sample can also be considered as constituting the filter shape.

[0569] There may be multiple samples whose values ​​are determined by filtering. Filter strength may represent the range of samples whose values ​​are determined by filtering. Filter strength may be either strong filtering strength or weak filtering strength. The number of samples whose values ​​are determined by strong filtering strength may be greater than the number of samples whose values ​​are determined by weak filtering strength. Alternatively, filter strength may represent the range of values ​​that are changed by filtering. The range of sample values ​​changed by strong filtering strength may be wider than the range of sample values ​​changed by weak filtering strength.

[0570] Filter coefficients can be coefficients or weights of the input samples.

[0571] The offset can be a specific value added to the result calculated using the values ​​and coefficients of the input samples, such as a weighted sum.

[0572] Filtering, interpolation, and sampling may be common in that they update the values ​​of samples. Accordingly, the description of any one of filtering, interpolation, and sampling in the embodiments may also apply to the other one of filtering, interpolation, and sampling. Here, sampling may include at least one of upsampling, downsampling, and subsampling.

[0573] Filtering may include filtering performed by a predictor (123) and a predictor (163), etc.

[0574] In encoding for a target block, a prediction error may exist between the original sample of the original block and the prediction sample of the prediction block. To reduce the prediction error, filtering may be performed on at least one of the prediction sample of the prediction block and the reference sample referenced for prediction.

[0575] For example, in intra-prediction, the reference samples may include one or more of the top-left reference sample, top reference sample, top-right reference sample, left reference sample, and bottom-left reference sample. Filtering of the prediction samples may be performed by applying specific weights to the prediction samples, left reference samples, top reference samples, and / or top-left reference samples, respectively.

[0576] Filtering for at least one of the prediction sample and the reference sample may be performed based on the attributes of the target block and the attributes of the prediction sample. For example, whether filtering is performed, the type of filter, the area to which filtering is applied, the weights of the filtering, the reference sample, the range of the reference sample, and the location of the reference sample may each be determined based on the attributes of the target block and the attributes of the prediction sample.

[0577] For example, the attributes of the target block may include information related to the target block described in the embodiments, such as 1) size, 2) prediction mode, 3) intra prediction mode, 4) reference sample line, 5) sample value, and 6) coding parameter.

[0578] For example, the attributes of the prediction sample may include information related to the prediction sample described in the embodiments, such as 1) the sample value and 2) the location within the target block of the prediction sample, and may include coding parameters regarding the prediction sample.

[0579] Filtering may include in-loop filtering performed by a filter (130) and a filter (170), etc.

[0580]

[0581] Figure 6 shows a plurality of in-loop filters according to one example.

[0582] Multiple in-loop filters of in-loop filtering may include one or more of Luma Mapping with Chroma Scaling (LMCS), deblocking filter, Sample Adaptive Offset (SAO), and Adaptive Loop Filter (ALF).

[0583] Multiple in-loop filters can be connected sequentially. For example, multiple in-loop filters can be connected in the order of LMCS, deblocking filter, SAO, and ALF. Additionally, multiple in-loop filters can be connected in any order of all available permutations of the multiple in-loop filters. The output from one of the multiple in-loop filters can be used as an input to the next filter.

[0584] As illustrated in FIG. 6, an input image may be input to the first filter. The input image may be a block as described in the embodiments. For example, the input image may be a restored block generated by an adder (129) or an adder (169). The output from one filter may be input to the next filter. An output image may be generated by the last filter. The output image may be a filtered block as described in the embodiments. For example, the output image may be a filtered restored image generated by a filter (130) or a filter (170).

[0585] The target block can represent the image input to the filter. The filtered target block can represent the image output from the filter.

[0586] LMCS may include luminance signal mapping for the luminance signal of the target block and chroma signal scaling for the chroma signal of the target block.

[0587] Luminous signal mapping can perform codeword redistribution for the luminous signal.

[0588] Luma signal mapping may include forward mapping and inverse mapping. In forward mapping, the existing dynamic range may be divided into multiple intervals. The mapped dynamic range can be determined by performing codeword redistribution on the input image using a linear model for each interval. In inverse mapping, inverse mapping from the mapped dynamic range to the existing dynamic range is performed.

[0589] Chroma scaling can correct the chroma signal based on the interrelationship between the luminance signal and the corresponding chroma signal.

[0590] Forward mapping can be performed between inter-prediction of the luminance signal and restoration of the luminance signal, and between inter-prediction of the luminance signal and chroma scaling. Inverse mapping can be performed between restoration of the luminance signal and in-loop filtering of the luminance signal. Chroma scaling can be performed between inverse transform and restoration of the chroma signal.

[0591] According to this structure, inverse quantizations for the luminance and chroma signals, inverse transforms for the luminance and chroma signals, prediction for the luminance signal, and restoration for the luminance signal can be performed within a mapped dynamic domain. In-loop filtering for the luminance and chroma signals, inter-predictions for the luminance and chroma signals, intra-predictions for the chroma signal, and restoration for the chroma signal can be performed within the existing dynamic domain.

[0592] A deblocking filter can remove block distortion occurring at the boundaries between blocks within the reconstructed image. For example, the blocks may be transformed blocks. Additionally, the blocks may be subblocks of a specific block described in the embodiments. Here, the boundaries between blocks may refer to samples adjacent to the boundaries between blocks.

[0593] A deblocking filter can be applied to the vertical and horizontal boundaries between blocks. After filtering is performed on the vertical boundaries of the blocks, filtering can be performed again on the horizontal boundaries of the filtered blocks.

[0594] A deblocking filter may be applied optionally. Whether to apply a deblocking filter to a target block may be determined based on at least one of sample(s) contained within a specific number of columns or rows within the target block and sample(s) contained within a specific number of columns or rows within a neighboring block adjacent to a specific boundary.

[0595] When a deblocking filter is applied to a target block, the filter to be applied may be determined according to the required strength of deblocking filtering. In other words, among a plurality of different filters, the filter determined according to the strength of deblocking filtering may be applied to the target block. The plurality of filters may include one of a long-tap filter, a strong filter, a weak filter, and a Gaussian filter.

[0596] The maximum length of the deblocking filter can be determined based on attributes of the target block, such as the size of the target block, the components of the target block, and coding parameters.

[0597] SAO can compensate for distortion between the original image and the reconstructed image on a sample basis. For compensation, SAO can apply an appropriate offset to the sample values. In other words, the offset can be added to the sample values.

[0598] An offset can be determined for the target block. For example, the offset can be determined for each component of the CTB. The determined offset can be applied to samples within a specific component of the CTB.

[0599] SAO may include an SAO using an edge offset (EO) and an SAO using a band offset (BO). Depending on the characteristics of samples within a specific block, such as a CTU, whether to perform an SAO using EO and whether to perform an SAO using BO may be determined, respectively.

[0600] In an SAO using EO, correction for distortion of samples can be performed based on the direction of edges within the target block. The pattern classes of the EO may include horizontal patterns, vertical patterns, 135-degree diagonal patterns, and 45-degree diagonal patterns. For the target block, information indicating the pattern class applied to the target block and multiple offsets of the said pattern class may be signaled. There may be four offsets. For a target sample within the target block, adjacent samples of the target sample may be determined according to the direction of the pattern class. An offset to be applied to the target sample may be determined by the pattern of the adjacent samples.

[0601] In an offset using BO, correction for sample distortion can be performed by classifying the brightness values ​​of samples within the target block into specific bands. The bit depth of the input image can be divided into m intervals. For example, m can be 32. The specific bands can be n consecutive intervals among the m intervals. For example, n can be 4. n offsets for the n intervals can be signaled. Additionally, information indicating the first interval selected as the n intervals among the m intervals can be signaled. The offset of the interval corresponding to the target sample can be added to the sample value of the target sample of the target unit.

[0602] ALF can compensate for distortion between the restored image and the original image.

[0603] The filter coefficients of the ALF can be signaled through the bitstream.

[0604] The filter shape of the ALF can be determined by the components of the target block. For example, a 7x7 diamond-shaped filter can be used for the luminance component. A 5x5 diamond-shaped filter can be used for the chroma component.

[0605] In ALF, the characteristics of a specific block can be determined for that block, and the class of that specific block can be determined based on those characteristics. In other words, the determination of characteristics and the determination of the class in ALF can be performed in units of 4x4 blocks. Filter coefficients can be calculated according to the class. A specific block can be a block with a size of 4x4.

[0606] One of 25 classes can be determined as the class of a specific block based on the direction and activity determined using the gradient of the specific block. Depending on the gradient of the specific block, a rotation transformation, a vertical reflection transformation, and / or a diagonal reflection transformation may be applied to the filter.

[0607] Information regarding whether ALF is applied can be signaled to specific units such as CTBs.

[0608] An index indicating a filter to be applied to a specific unit among the available filters may be signaled. Here, the available filters may include fixed filters and filters configured using a parameter set. For example, the parameter set may be an Adaptive Parameter Set (APS). The fixed filters may be predefined identically in the encoding device (110) and the decoder (150). The filter coefficients of the filters configured using the parameter set may be determined based on the coding parameters.

[0609]

[0610] Entropy Encoding and Entropy Decoding

[0611] Figure 7 shows entropy encoding and entropy decoding according to one example.

[0612] At the top of Fig. 7, the processes of entropy encoding by the entropy encoder (139) are illustrated.

[0613] The entropy encoder (139) may include a context modeler, a binarization unit, and an entropy encoding unit. The context modeler may include a context selection unit and a context memory.

[0614] The binarization unit can generate binaries for syntactic elements by performing binarization on the syntactic elements of the target block. Binarization may be a process of converting syntactic elements into the form of binaries.

[0615] Information about syntax elements and beans can be provided from the binarization unit to the context selection unit.

[0616] The context modeler can perform context updates.

[0617] Context can refer to occurrence probability information for each bin regarding syntactic elements that have already been encoded.

[0618] The context modeler may perform a context update to apply current probability information to the entropy encoding of the bins of the syntactic elements of the target block. The updated context may be stored in context memory. At this time, the updated context corresponding to the syntactic elements of the target block (or the bins within the syntactic elements of the target block) may be derived by the context modeler.

[0619] The context selector can select a context corresponding to a bin of a syntactic element of a target block. The selected context can be loaded from context memory and used as an updated context for entropy encoding of the bins of the syntactic element of the target block.

[0620] The updated context can be used for entropy encoding of syntactic elements of the target block.

[0621] The entropy encoding unit can generate encoded information for syntactic elements of a target block by performing entropy encoding using generated bins and an updated context, and can generate a bitstream containing the encoded information. The entropy encoding unit may use at least one of an arithmetic encoding method and a bypass encoding method.

[0622] At the bottom of Fig. 7, the processes of entropy decoding by the entropy decoder (161) are illustrated.

[0623] The entropy decoder (161) may include a context modeler, an entropy decoder, and an inverse binary converter. The context modeler may include a context selection unit and a context memory.

[0624] The context modeler can perform context updates.

[0625] Context can refer to probability information regarding the occurrence of each bin for syntactic elements that have already been decoded.

[0626] The context modeler may perform a context update to apply the currently decoded probability information to the entropy decoding for the bins of the syntactic elements of the target block. The updated context may be stored in context memory. At this time, the updated context corresponding to the syntactic elements of the target block (or the bins within the syntactic elements of the target block) may be derived by the context modeler.

[0627] The context selection unit can select a context corresponding to a bin of a syntactic element of a target block. The selected context can be loaded from context memory and can be used as an updated context for entropy decoding of the syntactic element of the target block.

[0628] The updated context can be used for entropy decoding of the syntactic elements of the target block.

[0629] The entropy decoding unit can generate bins for the segmentation elements of the target block by performing entropy decoding on the encoded information of the bitstream based on the updated context. The entropy decoding unit may use at least one of an arithmetic decoding method and a bypass decoding method.

[0630] The debinaryization unit can obtain syntactic elements of a target block by performing debinaryization on at least one of the generated beans. Debinaryization may be a process of converting at least one of the beans into the form of a syntactic element.

[0631] Information about syntax elements and beans can be provided from the inverse binary unit to the context selection unit.

[0632] The syntax element may be one of the coding parameters described in the examples.

[0633]

[0634] Methods for binarization, inbinarization, entropy encoding, and entropy decoding

[0635] In the embodiments, to perform signaling for specific information, one or more of the binarization method, inverse binarization method, entropy encoding method and entropy decoding method listed below may be used.

[0636] - Signed 0-th order Exponential Golomb binarization / debinarization method (abbreviated as se(v))

[0637] - Signed k-order exponential-Golomb binarization / debinarization method (abbreviated as sek(v))

[0638] - 0-order exponentiation-Golomb binarization / debinarization method for unsigned positive integers (abbreviated as ue(v))

[0639] - k-order exponential-Golomb binarization / debinarization method for unsigned positive integers (abbreviated as uek(v))

[0640] - Fixed-length binarization / debinarization method (abbreviated as f(n))

[0641] - Truncated Rice binarization / debinarization method or truncated unary binarization / debinarization method (abbreviated as tu(v))

[0642] - Truncated binary binarization / debinarization method (abbreviated as tb(v))

[0643] - Context-adaptive arithmetic encoding / decoding method (abbreviated as ae(v))

[0644] - Bit string in bytes (abbreviated as b(8))

[0645] - Signed integer binary / debinary conversion method (abbreviated as i(n))

[0646] - Unsigned positive integer binarization / debinarization method (abbreviated as u(n)) ('u(n)' may also refer to a fixed-length binarization / debinarization method.)

[0647] - Unary Binary / Debinary Method

[0648]

[0649] Prediction of embodiments

[0650] In the embodiments, a method, apparatus, and recording medium for image encoding / decoding are described.

[0651] The extrapolation filter described in the embodiments may be used for intra-prediction, but is not limited thereto, and the extrapolation filter or the description of the extrapolation filter in the embodiments may also be applied to other predictions described in the embodiments, such as inter-prediction, or other processing or other filters for blocks described in the embodiments, such as in-loop filters.

[0652] The derivation of candidates described in the embodiments may be used for Intra Template Matching Prediction (IntraTMP), but is not limited thereto, and the description of the candidates or the derivation of candidates in the embodiments may also be applied to other predictions using candidates, block vectors, or motion vectors, other processing of blocks, or other candidates.

[0653] In the embodiments, prediction methods may be fused. In addition to prediction methods such as intra-prediction and IntraTMP, other prediction methods described in the embodiments may be fused. In the embodiments, fusion may mean a weighted sum.

[0654] In one example, the fusion of prediction methods may mean that the prediction blocks generated by the prediction methods are fused. For instance, the fusion of prediction methods may mean that the weighted sum of the prediction blocks generated by the prediction methods is used as the (final) prediction block for the current block. Alternatively, the fusion of prediction methods may mean that the prediction blocks generated by the prediction methods are blended.

[0655] Prediction performance can be improved by the methods of the embodiments.

[0656] For example, when an intra-prediction technique based on an extrapolation filter is performed, filter values ​​from blocks encoded in the same mode can be inherited and used. However, since these filter values ​​are not optimized for the current block, there may be limitations to the prediction accuracy.

[0657] For example, in the prediction of chroma component blocks, there may be limitations to the prediction accuracy of the chroma block because the relationship between the encoding information and samples of the reconstructed luminance block at the same location and the surrounding samples of the current chroma block is utilized.

[0658] For example, when constructing block vector-based candidates, there may be limitations in constructing various block vector-based candidates if the search area is insufficient or if there are few blocks encoded based on block vectors.

[0659] In the embodiments, an extrapolation filter may be applied when chroma samples are predicted during image encoding / decoding. Additionally, in the embodiments, when intra-prediction of the current chroma block is performed during image encoding / decoding, the extrapolation filter used in the luminance block may also be applied to the chroma block. By applying these methods, prediction performance can be improved and encoding efficiency can be increased.

[0660] In the embodiments, during image encoding / decoding, an extrapolation filter for the current chroma block can be derived using decoded chroma samples surrounding the current chroma block. This decoding can improve the performance of the prediction technique.

[0661] In the embodiments, during image encoding / decoding, an extrapolation filter derived for the luminance component can be applied to the chroma component without deriving an additional chroma filter. By applying this, encoding overhead and computational load can be reduced.

[0662] In the embodiments, for encoding / decoding an image, the extrapolation filter of the current block can be derived from the decoding samples adjacent to the current block.

[0663] In the embodiments, during image encoding / decoding, filters of surrounding blocks to which extrapolation filters are applied may be inherited by the current block. Alternatively, a filter used by a luminance block at the same location may be used as a candidate for the current block. In this case, the inherited filter may be corrected to suit the characteristics of the current block, and the corrected filter may be used. The performance of the prediction technique may be improved by such usage.

[0664] In the embodiments, during image encoding / decoding, a candidate based on a block vector that can be derived from a P slice or B slice having a reference picture may be added as a candidate for the current block. Prediction performance may be improved by such addition.

[0665] In the embodiments, for image encoding / decoding, information on block vectors that can be derived through a P slice or a B slice can be used as a candidate based on a block vector for the current block within an I slice. Prediction performance can be improved by such use.

[0666] In the embodiments, during image encoding / decoding, a combination-based candidate for intra-prediction based on block vectors may be added as a candidate for the current block. Prediction performance may be improved by such addition.

[0667]

[0668] Image encoding using prediction

[0669] FIG. 8 is a flowchart of a method for predicting a current block and a method for generating a bitstream according to one embodiment.

[0670] The prediction method and bitstream generation method of the current block of the embodiment may be performed by an encoding device (110). The embodiment may be part of a encoding method of the current block or a video encoding method.

[0671] In step (810), the processor (120) can induce a prediction mode for the current block.

[0672] The prediction mode may be one of the modes using the prediction described in the embodiments.

[0673] The processor (120) can derive a prediction mode for the current block by considering the rate-distortion costs for the prediction modes among the prediction modes available for the current block.

[0674] For example, the prediction mode may be an intra prediction mode. The prediction for the current block may be an intra prediction mode, and an intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0675] For example, the processor (120) can derive an intra prediction mode for the current block by considering the rate-distortion costs for the intra prediction modes among the intra prediction modes available for the current block.

[0676] In step (820), the processor (120) can construct a reference sample for prediction.

[0677] The reference sample may be a reference sample used for the prediction described in the embodiments.

[0678] For example, the reference sample may be a reference sample in intra-prediction. The reference sample may be a reference sample used in a determined intra-prediction mode.

[0679] In step (830), the processor (120) can perform a prediction for the current block using an induced prediction mode. The processor (120) can generate a prediction block for the current block through the prediction.

[0680] For example, in step (830), the processor (120) can perform an intra prediction for the current block using an induced intra prediction mode. The processor (120) can generate a prediction block for the current block through the intra prediction.

[0681] Information about the encoded current block can be generated by performing a prediction on the current block using an induced prediction mode.

[0682] For example, information about the encoded current block can be generated by performing intra prediction on the current block using an induced intra prediction mode.

[0683] A prediction block can be generated by a prediction for the current block using an induced prediction mode, and a residual block, which is the difference between the current block and the prediction block, can be generated. Information about the encoded current block can be generated by applying transformation and quantization to the residual block.

[0684] For example, a prediction block can be generated by intra prediction for the current block using an induced intra prediction mode, and a residual block, which is the difference between the current block and the prediction block, can be generated. Information about the encoded current block can be generated by applying transformation and quantization to the residual block.

[0685] Information regarding the encoded current block may include transforms and quantized coefficients for the current block. Information regarding the encoded current block may include coding parameters for the current block.

[0686] In step (840), the processor (120) can generate a bitstream.

[0687] The bitstream may contain information about the currently encoded block.

[0688] The bitstream may contain prediction information.

[0689] The prediction information may be information regarding the prediction of the current block described above. The information regarding the prediction of the current block may include coding parameters related to the current block and / or sub-block, etc., for the prediction described in the embodiments.

[0690] For example, the prediction information may be information regarding the intra-prediction of the current block described above. The information regarding the intra-prediction of the current block may include coding parameters related to the current block and / or sub-block, etc., for the intra-prediction described in the embodiments.

[0691] Predicted information may be generated in step (840), or at least partially generated in steps (810, 820 and 830).

[0692] The processor (120) can store the generated bitstream in the storage (140). Alternatively, the communicator (149) can transmit the bitstream to the decoding device (150).

[0693] The processor (120) can perform entropy encoding for the prediction information and generate a bitstream containing the entropy-encoded prediction information.

[0694] An embodiment may be combined with the operation of the encoding device (110) described above with reference to FIG. 1. For example, the operations of step (810), step (820), and step (830) may be performed by a predictor (123). The operations of step (840) may be performed by an entropy encoder (139). Additionally, operations performed in other components of the encoding device (110) before, after, and between steps (810), step (820), step (830), and step (840) may be performed.

[0695]

[0696] Image decoding using prediction

[0697] FIG. 9 is a flowchart of a method for predicting a current block using a bitstream according to one embodiment.

[0698] The prediction method of the current block using the bitstream of the embodiment can be performed by a decoding device (150). The embodiment may be part of a decoding method of the current block or a video decoding method.

[0699] In step (910), the communicator (189) can acquire a bitstream. The communicator (189) can receive a bitstream from the encoding device (110).

[0700] The bitstream may contain information about the currently encoded block.

[0701] Information regarding the encoded current block may include transforms and quantized coefficients for the current block. Information regarding the encoded current block may include coding parameters for the current block.

[0702] The bitstream may contain prediction information.

[0703] The prediction information may be information regarding the prediction of the current block described above. The information regarding the prediction of the current block may include coding parameters related to the current block and / or sub-block, etc., for the prediction described in the embodiments.

[0704] For example, the prediction information may be information regarding the intra-prediction of the current block described above. The information regarding the intra-prediction of the current block may include coding parameters related to the current block and / or sub-block, etc., for the intra-prediction described in the embodiments.

[0705] The processor (160) can store the acquired bitstream in the storage (180).

[0706] The processor (160) can obtain prediction information from the bitstream. The processor (160) can obtain prediction information by performing entropy decoding on the entropy-encoded prediction information of the bitstream.

[0707] In step (920), the processor (160) can induce a prediction mode for the current block.

[0708] The prediction mode may be one of the modes using the prediction described in the embodiments.

[0709] For example, the prediction mode may be an intra prediction mode. The prediction for the current block may be an intra prediction mode, and an intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0710] The processor (160) can determine the prediction mode for the current block based on the prediction information.

[0711] For example, the prediction mode may be an intra prediction mode. The prediction for the current block may be an intra prediction mode, and an intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0712] For example, the processor (160) can determine an intra-prediction mode for the current block based on prediction information.

[0713] For example, the processor (160) can derive the prediction information for the current block from among the intra prediction modes available for the current block.

[0714] In step (930), the processor (160) can construct a reference sample for prediction.

[0715] The reference sample may be a reference sample used for the prediction described in the embodiments.

[0716] For example, the reference sample may be a reference sample in intra-prediction. The reference sample may be a reference sample used in a determined intra-prediction mode.

[0717] In step (940), the processor (160) can perform an intra prediction for the current block using an induced intra prediction mode. The processor (160) can generate a prediction block for the current block through the intra prediction.

[0718] In step (940), a prediction block can be generated by performing a prediction on the current block using the induced prediction mode, and a reconstructed block, which is the sum of the prediction block and the reconstructed residual block, can be generated.

[0719] For example, in step (940), a prediction block may be generated by performing an intra prediction on the current block using an induced intra prediction mode, and a restoration block which is the sum of the prediction block and the restoration residual block may be generated.

[0720] An embodiment may be combined with the operation of the decoding device (150) described above with reference to FIG. 1. For example, the operations of step (910) may be performed by an entropy decoder (161). The operations of steps (920), (930), and (940) may be performed by a predictor (163). Additionally, operations performed in other components of the decoding device (150) before, after, and between steps (910), (920), (930), and (940) may be performed.

[0721]

[0722] The necessity of efficient prediction and encoding / decoding methods for efficient prediction

[0723] In the embodiments, the image encoding / decoding method may use techniques such as histogram calculation and template matching for the reconstructed blocks / samples. However, these techniques may increase the complexity of encoding / decoding the image.

[0724] In the embodiments, the image encoding / decoding method may construct a histogram for the restored blocks / samples using a specified region. However, the improvement in the efficiency of encoding / decoding may be limited by the limitation of this region.

[0725] In the embodiments, the image encoding / decoding method may perform prediction for the luminance / chroma block using a weighted sum between a limited number of prediction blocks and may use a limited number of prediction candidates. However, the improvement in encoding / decoding efficiency may be limited by the use of such a limited number.

[0726] In the embodiments, the image encoding / decoding method may construct a candidate list for the current block using a prediction mode of a neighboring block. That is to say, the candidate list for the current block may be configured to include a prediction mode of a neighboring block.

[0727] In the embodiments, "neighboring block / sample / pixel" may refer to a block / sample / pixel within a specific area adjacent to the current block. The specific area may be an area specified for the current block described in the embodiments. Alternatively, "neighboring block / sample / pixel" may refer to a block / sample / pixel adjacent to the current block. Alternatively, "neighboring block / sample / pixel" may refer to a block / sample / pixel whose distance from the current block is less than or equal to a specific value.

[0728] For example, an image encoding / decoding method can construct a candidate list for the current block using the intra-prediction mode of the surrounding blocks. That is to say, the candidate list for the current block can be configured to include the intra-prediction mode of the surrounding blocks.

[0729] At this time, regarding the intra prediction of surrounding blocks, if an intra prediction technology other than a conventional intra prediction mode (e.g., directional mode, planar mode, and DC mode, etc.) is used, when constructing the candidate list of the current block, an alternative intra prediction mode corresponding to the other intra prediction technology may be used instead of the other intra prediction technology used for surrounding blocks.

[0730] That is, the candidate list of the current block may include an alternative intra prediction mode corresponding to the other intra prediction technology of the surrounding blocks instead of the other intra prediction technology. For example, this alternative prediction mode may be a fixed intra prediction mode. This alternative prediction mode may include a non-directional intra prediction mode such as a DC mode or a planner mode.

[0731] However, mapping to such a fixed intra-prediction mode may have limitations in improving encoding / decoding efficiency.

[0732] In the embodiments, the image encoding / decoding method may use multiple signalings for template matching prediction. The use of these multiple signalings may limit the improvement of encoding / decoding efficiency.

[0733] In the embodiments, the image encoding / decoding method can perform prediction using a fixed reference line. In this case, multiple signalings may be used so that a reference line among multiple available reference lines that is relatively far from the current block is used. Due to the use of these multiple signalings, the improvement of encoding / decoding efficiency may be limited.

[0734] In the embodiments, the image encoding / decoding method may use samples within a luminance block corresponding to the current block when performing prediction for a chroma block. The corresponding luminance block may be a block having the same location as the current block. By using samples within the corresponding luminance block, parallelization of prediction for the luminance block and prediction for the chroma block may be limited.

[0735] In the embodiments, the image encoding / decoding method can construct a histogram based on the amount of change in pixel values ​​of restored pixels within a fixed area and can induce an intra-prediction mode based on the histogram. Due to the use of such multiple signalings, the improvement in encoding / decoding efficiency may be limited.

[0736] The image encoding / decoding method described in the embodiments can overcome these limitations and may have a configuration to which these limitations do not apply.

[0737] In the embodiments, the image encoding / decoding method may extend the reference template construction step for analyzing a reference sample. Through this extension, the image can be efficiently encoded / decoded.

[0738] In the embodiments, the image encoding / decoding method can perform a prediction for a chroma component block using a plurality of pieces of information and can improve the prediction performance of the prediction.

[0739] In the embodiments, the image encoding / decoding method can derive / sort an intra prediction mode for analyzing a reference sample and can derive the intra prediction mode based on a template matching cost. The complexity and signaling overhead of these methods for deriving / sorting the intra prediction mode can be reduced.

[0740] In the embodiments, the image encoding / decoding method may extend the step of including surrounding blocks in a candidate list. Through this extension, the image can be efficiently encoded / decoded.

[0741] In the embodiments, the image encoding / decoding method can construct a candidate list of the current block by applying directionality to a non-directional mode when an intra prediction technique other than an intra prediction mode is applied to a surrounding block. Here, the intra prediction mode may include a directional mode, a planner mode, and a DC mode. Prediction performance can be improved by applying directionality to a non-directional mode.

[0742] In the embodiments, the image encoding / decoding method can perform prediction for chroma component blocks using block vector information. Prediction performance can be improved by using block vectors.

[0743] In the embodiments, the image encoding / decoding method can extend the reference line during the process of deriving an intra-prediction mode through the analysis of a reference sample. Through the extension of the reference line, the image can be efficiently encoded / decoded.

[0744] By the methods described above, signaling overhead can be reduced, and the time complexity required for prediction can be reduced.

[0745]

[0746] Prediction using an induced prediction mode

[0747] Figure 10 is a flowchart of a prediction method according to one example.

[0748] The prediction method of the embodiments can be performed by an encoding device (110) and / or a decoding device (150).

[0749] For example, the encoding device (110) can perform the intra-prediction method of the embodiment to compare the efficiencies of a plurality of prediction methods for the current block, and can perform the intra-prediction method of the embodiment to generate a reconstructed block for the current block.

[0750] In one embodiment, the current block may be a CU and may be one of the blocks described in the embodiments. For example, the current block may be at least one of a CTB, CU, PU, ​​TU, sub-block, a block having a specified block size, and a block within a predefined range of block sizes. Alternatively, the current block may represent a unit of coding. Alternatively, the current block may represent a specified area within a target picture.

[0751] In the following, the size of the current block may be WxH. W may represent the width of the current block. H may represent the height of the current block. Each of W and H may be a positive integer. Each of W and H is 2 n n can be a positive integer.

[0752] For example, the decoding device (150) can perform the intra-prediction method of the embodiment to generate a restoration block for the current block.

[0753] In the following, the processor may correspond to the processor (120) of the encoding device (110) and / or the processor (160) of the decoding device (150).

[0754] In step (1010), the processor can induce a prediction mode for the current block.

[0755] Step (1010) may correspond to Step (810) described with reference to FIG. 8 and Step (920) described with reference to FIG. 9. The content described in Step (810) described with reference to FIG. 8 or Step (920) described with reference to FIG. 9 may also be applied to Step (1010).

[0756] For example, the prediction mode may be a mode that uses the prediction described in the embodiments.

[0757] For example, the prediction mode may be the intra prediction mode used in intra prediction.

[0758] The prediction mode for the current block can be derived using an artificial neural network.

[0759] For example, the intra-prediction mode for the current block can be derived using an artificial neural network.

[0760] The processor includes: 1) an extrapolation filter-based intra prediction mode (EIP); 2) a method using the intra prediction mode of a neighboring block (e.g., Most Probable Mode (MPM)); 3) a method deriving the intra prediction mode using reference sample prediction (e.g., Template-based Intra Mode Derivation (TIMD)); 4) a method deriving the intra prediction mode using reference template / sample analysis (e.g., Decoder-side Intra Mode Derivation (DIMD)); and 5) an intra template matching mode deriving the intra prediction mode based on a template (e.g., Intra Template Matching Prediction (IntraTMP)). and 6) a method of deriving based on the intra prediction mode of the reference block (e.g., Occurrence-Based Intra Coding (OBIC)); one or more of these methods may be used to derive an intra prediction mode for the current block. Additionally, methods related to prediction and / or prediction modes described in the embodiments may be used for deriving / coding / decoding the intra prediction mode of the current block.

[0761] In the embodiments, the intra template matching mode can induce an intra prediction mode based on the template matching cost.

[0762] In the embodiments, the neighboring block may mean an adjacent block.

[0763] In the embodiments, the intra prediction mode candidate list may be a list of Most Probable Modes (MPMs). For example, the intra prediction mode for the current block may be derived using the MPM list.

[0764] The MPM list may contain one or more values. Each of the one or more values ​​may represent one or more intra prediction modes. In the following, for the MPM list, the "value" of the MPM list, the "MPM" of the MPM list, the "intra prediction mode" of the MPM list, and the "candidate" of the MPM list may be used interchangeably.

[0765] For example, the intra prediction mode may include a DC mode, a planar mode, an angular prediction mode, a Position Dependent Intra Prediction Combination (PDPC) mode, a Matrix-based Intra Prediction (MIP) mode, a Matrix-based Position-Dependent Intra Prediction (MPDIP) or Position Dependent Prediction (PDP) mode, and an LM mode (i.e., intra_fromLuma mode). In this case, the matrix-based intra prediction mode may refer to a mode that defines matrix coefficient or weight values ​​based on the block size / shape and the intra prediction mode, and applies the matrix to a reference sample to make a prediction. Here, the reference sample may be a value corrected based on at least one of the reference sample described in the embodiments, the mean of the reference sample, and a downsampled reference sample. For example, statistical values ​​described in the embodiments, such as the average, may be derived for a plurality of reference samples described in the embodiments, and these statistical values ​​may be used as (final) reference samples. For example, one or more upsampled / downsampled / filtered / interpolated reference samples may be generated by applying upsampling / downsampling / filtering / interpolation to a plurality of reference samples described in the embodiments, and the generated one or more reference samples may be used as (final) reference samples. Here, the types of reference samples to which correction is applied and the types of reference samples derived by correction may be different from each other. The type may be at least one of the types of reference samples described in the embodiments.

[0766] In intra prediction, when information regarding the intra prediction mode of the current block, such as an MPM list, is transmitted, the intra prediction modes with the highest probability of matching the current block's intra prediction mode can be selected as MPMs in the MPM list by referring to the information of the current block's surrounding blocks. Through this selection, if the intra prediction mode used for intra prediction of the current block belongs to an MPM within the MPM list, a smaller number of binary bits may be used.

[0767] When an intra prediction mode list for the target component of the current block is configured, the processor may use a method to derive an intra prediction mode for the target component of the current block using MPM, and may use a method to derive an intra prediction mode for the target component of the current block using 1) a specific intra prediction mode and 2) an intra prediction mode for the reference component of the current block.

[0768] For example, the target component may be a chroma component. The reference component may be a luma component. Or, the target component may be a luma component. The reference component may be a chroma component.

[0769] For example, the target component may be one of the three or four components mentioned above, and the reference component may be another of the three or four components mentioned above.

[0770] When the intra prediction mode of the current block is derived using the MPM list, the processor may reorder the MPMs in the MPM list using 1) a context-adaptive MPM reordering method that uses information from surrounding blocks and / or 2) a context-non-adaptive MPM reordering method.

[0771] Here, the context-adaptive MPM reordering method can predict 1) candidates for the MPM list and 2) a reordering method for the said candidates by using at least one of the information of the previously restored surrounding block and the information of the current block.

[0772] Through such prediction, an intra-prediction mode for the current block can be directly induced using at least one of the information of the previously restored surrounding blocks and the information of the current block, without signaling of information regarding the prediction mode for the current block.

[0773] In step (1020), the processor can construct a reference sample for the prediction of the current block.

[0774] Step (1020) may correspond to Step (820) described with reference to FIG. 8 and Step (930) described with reference to FIG. 9. The content described in Step (820) described with reference to FIG. 8 or Step (930) described with reference to FIG. 9 may also be applied to Step (1020).

[0775] The processor may construct a reference sample using one or more of the following methods: 1) selection of a reference sample; 2) filtering of the reference sample; 3) selection of multiple reference sample lines; and 4) construction of a reference sample for sample analysis.

[0776] Additionally, methods related to the reference sample in the embodiments may be used for the construction of the reference sample. For example, methods related to filtering / sampling / interpolation for the sample described in the embodiments may be applied to the reference sample.

[0777] In step (1030), the processor can perform a prediction for the current block using the induced prediction mode.

[0778] Step (1030) may correspond to Step (830) described with reference to FIG. 8 and Step (940) described with reference to FIG. 9. The content described in Step (830) described with reference to FIG. 8 or Step (940) described with reference to FIG. 9 may also be applied to Step (1030).

[0779] The processor can perform a prediction for the current block using an induced prediction mode. Alternatively, the processor can perform an intra prediction for the target component of the current block using an induced intra prediction mode.

[0780] The processor can generate a predicted block for the current block through a prediction for the current block.

[0781] In one embodiment, the processor comprises: 1) non-directional intra prediction; 2) directional intra prediction; 3) inter-color prediction; 4) prediction for units of sub-blocks; 5) template matching-based prediction (e.g., Template Matching Prediction (TMP)); 6) prediction using intra prediction fusion; 7) prediction using an intra partitioning prediction mode (e.g., Spatial Geometric Partitioning Mode (SGPM)); 8) intra prediction using a template-based multiple reference line (e.g., Template-based Multiple Reference Line (TMRL)); and 9) prediction using an extrapolation filter-based intra prediction mode (e.g., Extrapolation filter-based Intra Prediction (EIP) mode); 10) Intra prediction for the current block can be performed using one or more matrix-based intra prediction methods.

[0782] Additionally, the processor may perform corrections on the samples of the prediction. The prediction may be an intra-prediction. Methods related to filtering / sampling / interpolation for samples described in the embodiments may be applied to the prediction samples.

[0783] The processor can perform filtering on the prediction samples of the prediction block during the process of performing intra-prediction.

[0784]

[0785] Encoding method applying an extrapolation filter

[0786] Figure 11 is a flowchart of an encoding method using an extrapolation filter according to one example.

[0787] Below, an encoding method using an extrapolation filter performed by an encoding device (110) is described. That is to say, the case in which an extrapolation filter is used in the prediction method of the current block and the bitstream generation method described above with reference to FIG. 8 is described in more detail.

[0788] In step (1110), an extrapolation filter for the current block can be configured.

[0789] In one example, an extrapolation filter for the current block can be derived using the coding parameters of the embodiments and samples within the restored region surrounding the current block.

[0790] In one example, the extrapolation filter can be inherited. For example, the specific extrapolation filter of the restored block on which a prediction using a specific extrapolation filter was performed can be inherited by the current block.

[0791] For example, a specific extrapolation filter of the restored block can be inherited as is by the current block.

[0792] For example, specific filter information of a restored block on which a prediction using a specific extrapolation filter was performed may be inherited by the current block. The specific filter information may be information constituting the extrapolation filter described in the embodiments or information used to constitu the extrapolation filter.

[0793] In one example, an extrapolation filter of a specific object may be inherited by the current block. The specific object may be one of the objects, units of processing, and objects of processing described in the embodiments. The specific object may be an object associated with the block described in the embodiments. For example, the specific object may be an object containing the block. For example, the specific object may be a specific location within the block.

[0794] For example, a specific extrapolation filter of a specific object can be inherited as is into the current block.

[0795] For example, specific filter information of a specific object on which a prediction using a specific extrapolation filter was performed may be inherited by the current block. The specific filter information may be information constituting the extrapolation filter described in the embodiments or information used to constitu the extrapolation filter.

[0796] In one example, the extrapolation filter may be a filter selected for prediction for the current block, or a candidate filter configured for prediction for the current block.

[0797] The coefficient values ​​of these selected filters and / or candidate filters can be corrected, and correction coefficients can be derived.

[0798] In one example, values ​​related to the filter described in the embodiments, such as coding parameters, can be derived.

[0799] In step (1120), a prediction for the current block using an extrapolation filter can be performed.

[0800] In one example, for the prediction of the current block, reference samples described in the embodiments may be used. For example, the reference samples may include the restored surrounding samples described in the embodiments; and prediction samples of the current block. The prediction block may be generated and restored using the reference samples.

[0801] In one example, the prediction block may be corrected using an extrapolation filter. Alternatively, the prediction block may be corrected using the filter described in the embodiments. That is to say, the description of the correction of the prediction block using the extrapolation filter described in the embodiments may also be applied to other filters.

[0802] In step (1120), prediction mode information may be generated. Prediction mode information indicating the prediction mode generated for encoding the current block by the processing unit (110) of the encoding device (110) may be generated.

[0803] The prediction mode information may include information related to the prediction mode described in the embodiments.

[0804] For example, the prediction mode information may include information indicating whether an extrapolation filter-based prediction mode is used. For example, the prediction mode information may include information used to determine whether an extrapolation filter-based prediction mode is used.

[0805] In one example, the prediction mode information may include information for configuring an extrapolation filter.

[0806] For example, the prediction mode information may include information indicating the shape and size of the extrapolation filter.

[0807] For example, the prediction mode information may include information indicating a method for designing an extrapolation filter.

[0808] For example, prediction mode information may include information used in the fusion of prediction methods or predictors. The prediction mode information may include information deriving / specifying the prediction methods to be fused; information deriving / specifying the predictors to be fused; and weights for the prediction methods to be fused or the predictors to be fused.

[0809] For example, the prediction mode information may include correction information for the extrapolation filter. The correction information may include information used for correction of the filter described in the embodiments.

[0810] For example, the prediction mode information may include the values; information; and coding parameters described in the embodiments.

[0811] In step (1130), encoding of the prediction mode information can be performed.

[0812] Predicted mode information encoded by encoding of predicted mode information can be generated.

[0813] The bitstream may include (encoded) extrapolation filter-based prediction mode information.

[0814] The described steps (1110, 1120, and 1130) may be part of other steps described in the embodiments. For example, with reference to FIG. 8, the aforementioned step (830) may include step (1110) and step (1120). Step (840) may include step (1130).

[0815]

[0816] Decoding method applying an extrapolation filter

[0817] Figure 12 is a flowchart of a decoding method using an extrapolation filter according to one example.

[0818] Below, a decoding method using an extrapolation filter performed by a decoding device (150) is described. That is to say, the case in which an extrapolation filter is used in the prediction method of the current block using the bitstream described above with reference to FIG. 9 is described in more detail.

[0819] The bitstream may include (encoded) extrapolation filter-based prediction mode information.

[0820] In step (1210), decoding of the encoded prediction mode information can be performed.

[0821] Prediction mode information can be generated by decoding the encoded prediction mode information.

[0822] In step (1210), prediction mode information can be obtained. Prediction mode information indicating the prediction mode used for decoding the current block can be generated by the processing unit (160) of the decoding device (150).

[0823] The prediction mode information may include information related to the prediction mode described in the embodiments.

[0824] For example, the prediction mode information may include information indicating whether an extrapolation filter-based prediction mode is used. For example, the prediction mode information may include information used to determine whether an extrapolation filter-based prediction mode is used.

[0825] In one example, the prediction mode information may include information for configuring an extrapolation filter.

[0826] For example, the prediction mode information may include information indicating the shape and size of the extrapolation filter.

[0827] For example, the prediction mode information may include information indicating a method for designing an extrapolation filter.

[0828] For example, prediction mode information may include information used in the fusion of prediction methods or predictors. The prediction mode information may include information deriving / specifying the prediction methods to be fused; information deriving / specifying the predictors to be fused; and weights for the prediction methods to be fused or the predictors to be fused.

[0829] For example, the prediction mode information may include correction information for the extrapolation filter. The correction information may include information used for correction of the filter described in the embodiments.

[0830] For example, the prediction mode information may include the values; information; and coding parameters described in the embodiments.

[0831] In step (1220), an extrapolation filter for the current block can be configured.

[0832] In one example, an extrapolation filter for the current block can be derived using the prediction mode information of the embodiments, coding parameters, and samples within the restored region surrounding the current block.

[0833] In one example, the extrapolation filter can be inherited. For example, the specific extrapolation filter of the restored block on which a prediction using a specific extrapolation filter was performed can be inherited by the current block.

[0834] For example, a specific extrapolation filter of the restored block can be inherited as is by the current block.

[0835] For example, specific filter information of a restored block on which a prediction using a specific extrapolation filter was performed may be inherited by the current block. The specific filter information may be information constituting the extrapolation filter described in the embodiments or information used to constitu the extrapolation filter.

[0836] In one example, an extrapolation filter of a specific object may be inherited by the current block. The specific object may be one of the objects, units of processing, and objects of processing described in the embodiments. The specific object may be an object associated with the block described in the embodiments. For example, the specific object may be an object containing the block. For example, the specific object may be a specific location within the block.

[0837] For example, a specific extrapolation filter of a specific object can be inherited as is into the current block.

[0838] For example, specific filter information of a specific object on which a prediction using a specific extrapolation filter was performed may be inherited by the current block. The specific filter information may be information constituting the extrapolation filter described in the embodiments or information used to constitu the extrapolation filter.

[0839] In one example, the extrapolation filter may be a filter selected for prediction for the current block, or a candidate filter configured for prediction for the current block.

[0840] The coefficient values ​​of these selected filters and / or candidate filters can be corrected, and correction coefficients can be derived.

[0841] In one example, values ​​related to the filter described in the embodiments, such as coding parameters, can be derived.

[0842] In step (1230), a prediction for the current block using an extrapolation filter can be performed.

[0843] In one example, for the prediction of the current block, reference samples described in the embodiments may be used. For example, the reference samples may include the restored surrounding samples described in the embodiments; and prediction samples of the current block. The prediction block may be generated and restored using the reference samples.

[0844] In one example, the prediction block may be corrected using an extrapolation filter. Alternatively, the prediction block may be corrected using the filter described in the embodiments. That is to say, the description of the correction of the prediction block using the extrapolation filter described in the embodiments may also be applied to other filters.

[0845] The described steps (1210, 1220, and 1230) may be part of other steps described in the embodiments. For example, with reference to FIG. 9, the aforementioned step (910) may include step (1210). Step (940) may include step (1220) and step (1230).

[0846]

[0847] Intra prediction applying extrapolation filters

[0848] Below, a prediction using an extrapolation filter is described in detail. The descriptions may be considered as part of steps (1120) and (1230).

[0849] The prediction mode of the embodiments may include an extrapolation filter-based prediction mode that performs prediction based on an extrapolation filter.

[0850] Extrapolation filter-based prediction modes can include EIP modes.

[0851] In the embodiments, EIP may mean extrapolation filter-based intra prediction using the extrapolation filter described in the embodiments.

[0852] EIP can use an N-tap filter-based prediction method.

[0853] The shape of the EIP filter can be adaptively determined based on the information of the predicted block.

[0854] Here, the form may refer to the attributes of the block described in the embodiments and may include size, shape position and coding parameters, etc.

[0855] Multiple filter shapes can be included in the candidate list.

[0856] A filter combination may represent multiple filters. Here, the multiple filters may be filters used for the prediction of the current block. Alternatively, the multiple filters may be filters within a candidate list. Alternatively, the multiple filters may be filters selected for the current block from among specific filters, such as those within the candidate list.

[0857] Multiple extrapolation filters can be applied in parallel.

[0858]

[0859] Coefficients of the extrapolation filter

[0860] An extrapolation filter can be composed of coefficients for multiple different purposes (or types).

[0861] The types of coefficients of the extrapolation filter may include the types described below. The filter combination may include at least one type of coefficient:

[0862] [Type 1]

[0863] Spatial coefficients: Spatial coefficients may represent coefficients that can emphasize or reduce samples at specific spatial locations for operations on restored or predicted sample values.

[0864] [Type 2]

[0865] Normalization Factor: A normalization factor can refer to a coefficient that limits predicted values ​​to a specific range.

[0866] [Type 3]

[0867] Brightness correction factor: The brightness correction factor refers to a factor that can improve prediction performance by taking into account the brightness difference of the image.

[0868] [Type 4]

[0869] Weighting Factor: A weighting factor can refer to a coefficient that adjusts the sample weights for samples within a specific area or adjusts the weights of a specific type of coefficient.

[0870] The extrapolation filter can be derived using restored samples around the current block.

[0871] An extrapolation filter can be derived through samples of the reference region pointed to by the derived block vector and the surrounding region of said reference region.

[0872] The extrapolation filter can inherit filter information from blocks encoded / decoded in a mode that uses the previous extrapolation filter.

[0873] In the embodiments, the filter information may be information constituting the extrapolation filter described in the embodiments or information used to constituting the extrapolation filter.

[0874] The extrapolation filter may inherit filter coefficients used within the object associated with the block that has been encoded / decoded. For example, the object associated with the block may include the units and frames described in the embodiments. The units may include parameter sets, headers, bricks, slices, CTUs, CUs, PUs, and TUs. The units may include VPSs, SPSs, PPSs, APSs, and DPSs. The units may include picture headers, subpicture headers, slice headers, tile group headers, and tile headers. The frames may include previous frames, current frames, next frames, reference frames, etc.

[0875]

[0876] Form of extrapolation filter

[0877] FIGS. 13a to 13c show the forms of extrapolation filters according to one example.

[0878] As illustrated in FIGS. 13a to 13c, the extrapolation filter can have various forms.

[0879] In FIGS. 13a through 13c, a white square may represent a prediction block that is the subject of the prediction. Alternatively, a white square may represent a prediction sample. Alternatively, a white square may represent a specific sample of the prediction block that is the subject of the prediction. For example, a specific sample may include a top-leftmost sample within the prediction block.

[0880] In FIGS. 13a to 13c, the shaded rectangles may represent the locations of samples used for operations for white rectangles. The shaded rectangles may represent the locations of samples used for filtering using an extrapolation filter.

[0881] In FIGS. 13a to 13c, the locations of 14 samples are exemplified.

[0882] Similar to FIGS. 13a to 13c, the shape of the extrapolation filter can be configured in various ways, such as a square shape, a shape that is vertically or horizontally extended, and a diamond shape.

[0883] In configuring the form of the extrapolation filter, an area that extends the surrounding area of ​​the predicted location may be added as the filter area. For example, a sample / area adjacent to the sample / area of ​​the extrapolation filter described in the example may be added as the sample / area of ​​the extrapolation filter.

[0884] The description of a specific filter disclosed in the embodiments may also be applied to an extrapolation filter. For example, the description of the coefficients of a specific filter; the information used to derive the filter; the derivation of the filter information; inheritance; and the form may also be applied to an extrapolation filter.

[0885]

[0886] Characteristics of prediction performance using extrapolation filters

[0887] When prediction based on an extrapolation filter is performed, one or more filters may be used.

[0888] When prediction based on an extrapolation filter is performed, two or more filters configured with the same shape may be used.

[0889] When prediction based on extrapolation filters is performed, two or more filters composed of different shapes may be used.

[0890] The prediction mode based on the extrapolation filter of the embodiments may include intra-prediction.

[0891] The prediction mode based on the extrapolation filter of the embodiments can be applied as an intra prediction of a prediction mode that performs prediction using a combination of inter prediction and intra prediction.

[0892] For example, prediction modes using such combinations may include Combined Inter Intra Prediction (CIIP), Geometric Partitioning Mode (GPM), and Geometric Partitioning Mode - Intra Block Copy (IBC-GPM), which are modes that fuse intra predictors and inter predictors.

[0893] In addition, the prediction mode based on the extrapolation filter of the embodiments may be applied to all prediction modes using an intra predictor. Alternatively, the prediction mode based on the extrapolation filter of the embodiments may replace the predictor of the embodiments.

[0894] This prediction mode may be one of the prediction-related modes described in the embodiments.

[0895]

[0896] Intra prediction with extrapolation filter applied

[0897] In the embodiments, a method, apparatus, and a recording medium storing a bit stream that can be used in at least one of the processes for performing intra prediction when video encoding / decoding using an intra prediction technique or an inter prediction technique is performed may be described.

[0898] Here, the process of performing intra-prediction may include an extrapolation filter-based intra-prediction mode.

[0899] The intra prediction techniques of the embodiments may include non-directional prediction modes (e.g., DC mode, PLANAR mode, and block vector-based prediction); directional prediction modes; chroma component prediction modes; sub-block unit prediction modes; intra-partitioning prediction modes (e.g., Spatial Geometric Partitioning Mode (SGPM)); intra-prediction mode derivation using reference sample prediction (e.g., Template-based Intra Mode Derivation (TIMD)); and Intra Block Copy (IBC) modes; etc.

[0900] Additionally, the intra-prediction technology of the embodiments may include technology that includes / uses the intra-prediction described in the embodiments.

[0901] The IBC modes of the embodiments may include IBC-GPM; and IBC-CIIP modes; etc.

[0902] The inter-prediction technology of the embodiments may include CIIP mode and GPM.

[0903] In an intra-prediction mode where an extrapolation filter is applied, such as in EIP mode, predictions for the current block can proceed from the top-left to the bottom-right of the current block. In other words, predictions can be applied to the prediction samples of the current block in the order from the top-left position to the bottom-right position. For these predictions, an extrapolation filter can be applied to the surrounding reconstructed samples and the predicted samples.

[0904] To generate predictors, an extrapolation filter may be applied once, and multiple extrapolation filters may be applied two or more times.

[0905] An N-tap filter may be used for filtering prediction blocks or prediction samples, and one or more of the following [Equation 1] through [Equation 4] may be used adaptively. N may be an integer greater than or equal to 2.

[0906] [Formula 1]

[0907] Pred(x, y) = ∑(C i × t(x - offsetX i , y - offsetY i )) + C N-1 × 2 bit depth - 1

[0908] [Equation 2]

[0909] Pred(x, y) = C N-1 (∑(C i × t(x - offsetX i , y - offsetY i ))) + C N-2 × 2 bit depth - 1

[0910] [Equation 3]

[0911] Pred(x, y) = ∑(C i × t(x - offsetX i , y - offsetY i ))

[0912] [Equation 4]

[0913] Pred(x, y) = C N-1 (∑(C i × t(x - offsetX i , y - offsetY i )))

[0914] Pred(x, y) can be the predicted value at position (x, y) within the current block.

[0915] In [Equation 1] through [Equation 4], i can take one of the values ​​from 0 to N-2 in order by ∑. That is to say, ∑ can represent the sum of values ​​from 0 to N-2.

[0916] For example, in the case of a 15-tab filter where N is 15, [Equation 1] can be considered as [Equation 1-1] below.

[0917] [Equation 1-1]

[0918]

[0919] C i can be a filter coefficient. i can be a value greater than or equal to 0 and less than or equal to N-1.

[0920] t(x - offsetX i , y - offsetY i ) can be restoration samples or prediction samples.

[0921] Each of offsetX and offsetY can have an integer value greater than or equal to 1.

[0922] The predicted sample values ​​can be clipped to the reference sample range rather than the entire sample value range.

[0923] Here, the reference sample region used to determine the reference sample range may be the same as the region used when generating filter coefficients.

[0924] When an extrapolation filter is applied, a predetermined sample unit can be determined. The extrapolation filter can be applied to each determined sample unit.

[0925] For example, samples located diagonally within the current block may be determined as the aforementioned predetermined sample units. These predetermined sample units or diagonally located samples may be determined differently depending on the attributes of the current block, such as the size or shape of the current block.

[0926] Here, the application of the extrapolation filter and the calculation of the predicted value for samples included within a predetermined sample unit can be processed simultaneously in parallel.

[0927] The extrapolation filter for the prediction mode of the embodiments can be derived from restored samples around the current block and can be inherited from a block previously encoded / decoded in this prediction mode.

[0928] In one embodiment, the shape of the extrapolation filter can be fixed.

[0929] In one embodiment, the form of the extrapolation filter may be determined based on the current block; the predicted block; the surrounding blocks; and / or the information / attributes of the objects associated with the blocks.

[0930] For example, the shape of the filter may include information for configuring the filter, such as the size of the filter and the number of filter coefficients.

[0931] For example, the attributes of a block may include size, shape, position, and coding parameters. For example, the block's position may indicate the location of the block's top-to-leftmost sample. Alternatively, the block's position may indicate the block's location within the frame.

[0932] For example, the block information may include block partitioning information, a reference region of the block, and a reference region available to the block. For example, coding parameters may include coding parameters described in embodiments, such as quantization parameters.

[0933] For example, an object associated with a block may include units and frames described in the embodiments. A unit may include a parameter set, a header, a brick, a slice, a CTU, a CU, a PU, and a TU. A unit may include a VPS, an SPS, a PPS, an APS, and a DPS. A unit may include a picture header, a subpicture header, a slice header, a tile group header, and a tile header. A frame may include a previous frame, a current frame, a next frame, a reference frame, etc. For example, information about an object may include a slice type, a tile type, and a picture type, etc., and may include coding parameters for the object.

[0934] Using the information described above, the shape and size of the extrapolation filter can be adaptively determined, and the number of candidates for the extrapolation filter can be determined. In one example, the number of candidates for the filter shape can be determined based on the size of the current block.

[0935] For example, if the current block size is smaller than a specific value, K filter shapes can be used as candidates. If the current block size is greater than or equal to a specific value, S shapes can be used as candidates. For example, K and S can be greater than or equal to 1, and S can be greater than K. Both K and S can be natural numbers.

[0936] For example, if the current block size is smaller than a specific value, an A-tab filter may be used, and if the current block size is greater than or equal to a specific value, a B-tab filter may be used. For example, A can be greater than 1, and B can be greater than A. Both A and B can be natural numbers.

[0937] Predictions can be performed using two or more extrapolation filters based on predefined values. For example, a Multi-Model (MM) EIP can be used.

[0938]

[0939] Applying extrapolation filters for chroma block intra prediction

[0940] FIG. 14 shows specific locations within a luminance block and a color difference block that is the subject of prediction according to one example.

[0941] In FIG. 14, corresponding Luma blocks and Chroma blocks are illustrated. The block on the left in FIG. 14 represents the Luma block. The block on the right in FIG. 14 represents the Chroma block.

[0942] In Fig. 14, specific locations TL, TR, C, BL, and BR within the Luma block are shown.

[0943] TL may be the top-leftmost sample within the luminance block. TR may be the top-rightmost sample within the luminance block. C may be the center sample within the luminance block. BL may be the bottom-leftmost sample within the luminance block. BR may be the bottom-rightmost sample within the luminance block.

[0944] The Luma block can be a restoration block. The Chroma block can be a prediction block that is the subject of prediction.

[0945]

[0946] Derivation of extrapolation filters for chroma component blocks

[0947] When an extrapolation filter is derived for a chroma component block, the extrapolation filter used to predict the lumina component block corresponding to the chroma component block can be inherited.

[0948] Here, an extrapolation filter can be derived based on information indicating how the luminal component block and the chroma component block are partitioned within the same CTU.

[0949] For example, if a single tree type (SINGLE_TREE) in which the luminance component and the chroma component are equally partitioned within the same CTU is used, and the current block is a chroma component block and is encoded / decoded in an extrapolation filter-based prediction mode, the extrapolation filter of the chroma component block can be inherited from the extrapolation filter used for prediction of the corresponding luminance component block.

[0950] For example, when a tree type (DUAL_TREE_LUMA or DUAL_TREE_CHROMA) for a dual tree structure in which the luminance component and the chroma component are partitioned independently within the same CTU is used, and the current block is a chroma component block and is encoded / decoded in an extrapolation filter-based prediction mode, the extrapolation filter of the chroma component block can be derived by one of the methods described below.

[0951] An extrapolation filter used for a sample at a specific location within a luminance block corresponding to a chroma component block can be inherited as an extrapolation filter of the chroma component block.

[0952] For example, a specific location is where the position of the top-left sample of the corresponding luma block is (x0, y0) and the horizontal length is W l and the vertical length is H l It may be. In this case, the specific location is, as illustrated in FIG. 14, the top-leftmost location (e.g., TL (x0, y0) in FIG. 14), the top-rightmost location (e.g., TR (x0+W) in FIG. 14). l-1, y0)), center position (e.g., C (x0+W) in Fig. 14) l / 2, y0+H l / 2)), bottom-leftmost position (e.g., BL (x0, y0+H) in Fig. 14 l -1)) and bottom-to-right position (e.g., BR (x0+W) in Fig. 14 l -1, y0+H l It could be one of -1)).

[0953] For example, the above locations may be sorted according to a predetermined priority. Depending on the order of the sorted locations, an extrapolation filter of a block containing a specific location may be identified, and the identified extrapolation filter may be inherited for a chroma block. To identify the extrapolation filter, prediction mode information associated with the extrapolation filter may be identified.

[0954] In one embodiment, an extrapolation filter may be derived based on one or more of: a sample surrounding the luminance block corresponding to the current chroma component block; a sample within the luminance block; a Cr sample surrounding the current chroma component block; and a Cb sample surrounding the current chroma component block. The surrounding sample may refer to a reference sample.

[0955] An extrapolation filter can be derived using the surrounding reference samples of the current chroma block.

[0956] For example, an extrapolation filter can be derived based on a Cr reference sample or a Cb reference sample. Additionally, an extrapolation filter can be derived using both the Cr reference sample and the Cb reference sample.

[0957] For example, an extrapolation filter can be derived using samples within a corresponding luma block that has already been restored.

[0958] For example, an extrapolation filter can be derived using both samples within the corresponding luminance block that have already been restored and reference samples around the chroma block.

[0959] For example, an extrapolation filter can be derived based on the surrounding samples of the luminance block, the samples within the luminance block, and the surrounding samples of the chroma block.

[0960]

[0961] Size of the extrapolation filter and the positions of the coefficients

[0962] FIGS. 15a to 15y show the positions of the size and coefficients of an extrapolation filter according to one example.

[0963] As illustrated in FIGS. 15a to 15y, the extrapolation filter can have various sizes and coefficients.

[0964] In FIGS. 15a through 15y, a white square may represent a prediction block to which an extrapolation filter is to be applied. Alternatively, a white square may represent a prediction sample to which an extrapolation filter is to be applied. Alternatively, a white square may represent a specific sample of a prediction block to which an extrapolation filter is to be applied. For example, a specific sample may include the top-leftmost sample within the prediction block.

[0965] As shown in Figs. 15p and 15u, if it appears that there are multiple white squares, the white square at the far right and bottom may represent a predicted sample.

[0966] In FIGS. 15a through 15y, the shaded rectangles may represent the coefficients of the extrapolation filter. The shaded rectangles may represent the pixel locations of reference samples referenced for filtering using the extrapolation filter.

[0967] In FIGS. 15a through 15y, the locations of 3, 4, 5, 6, 7, 8, 14, or 15 reference samples are respectively exemplified. That is to say, the extrapolation filter of the embodiments may be a 4-tab, 5-tab, 6-tab, 7-tab, 8-tab, 9-tab, 15-tab, or 16-tab filter.

[0968] The shaded squares can represent the inputs of the extrapolation filter. The white squares can represent the outputs of the extrapolation filter.

[0969] Below, the sizes and positions of the coefficients of the N-tap extrapolation filters for the luminance block and / or chroma block of the embodiments are described.

[0970] In one embodiment, the form of the extrapolation filter may be determined based on the current block; the predicted block; the surrounding blocks; and / or the information / attributes of the objects associated with the blocks.

[0971] For example, the shape of the filter may include information for configuring the filter, such as the size of the filter and the number of filter coefficients.

[0972] For example, the attributes of a block may include size, shape, position, and coding parameters. For example, the block's position may indicate the location of the block's top-to-leftmost sample. Alternatively, the block's position may indicate the block's location within the frame.

[0973] For example, the block information may include block partitioning information, a reference region of the block, and a reference region available to the block. For example, coding parameters may include coding parameters described in embodiments, such as quantization parameters.

[0974] For example, objects associated with a block may include units and frames described in the embodiments. Units may include parameter sets, headers, bricks, slices, CTUs, CUs, PUs, and TUs. Units may include VPSs, SPSs, PPSs, APSs, and DPSs. Units may include picture headers, subpicture headers, slice headers, tile group headers, and tile headers. Frames may include previous frames, current frames, next frames, reference frames, etc.

[0975] Using the information described above, the shape and size of the extrapolation filter can be adaptively determined, and the number of candidates for the extrapolation filter can be determined.

[0976] When an N-tap extrapolation filter is configured, among the N coefficients, at least two values ​​may be included: a coefficient value for a specific pixel location, an offset value (so to speak, bias), and a scaling value (so to speak, scaling factor).

[0977] Here, the coefficient value for a specific pixel location may refer to a coefficient that is multiplied by the restored or predicted sample values ​​around the current block.

[0978] Here, the offset value may be a coefficient multiplied by 2 bit depth-1, and may mean a coefficient for correcting the average error of the extrapolated value.

[0979] Here, the scaling value may refer to a coefficient for correcting filter strength or adjusting extrapolation strength.

[0980] For example, an N-tap filter can be configured to include N-2 coefficient values ​​for N-2 specific pixel locations, one coefficient value related to the offset value, and one coefficient for the scaling value.

[0981] For example, an N-tap filter can be configured to include N-1 coefficient values ​​for N-1 specific pixel locations and 1 coefficient for a scale value.

[0982] In the configuration of the N-tap filter, specific pixel locations can be determined as shown in each of FIGS. 15a to 15y. An extrapolation filter of the shape having 3 to 8 specific pixel locations as shown in FIGS. 15a to 15y may be used.

[0983]

[0984] Correction of extrapolation filters

[0985] Below, an example is described in which a filter corrected according to the attributes / information of the current block is used.

[0986] Such corrected filters may be used when the extrapolation filter of the luminance block or the extrapolation filter of the surrounding block of the chroma block is inherited and used by the merge technique; and when the extrapolation filter derived by the reference region surrounding the restored luminance block or chroma block is used; etc.

[0987] The corrections of the embodiments can be applied not only to the extrapolation filter of the chroma block but also to the correction of the extrapolation filter of the lumina block.

[0988] In the embodiments, the attributes of the block may include size, shape, position, and coding parameters, etc. For example, the block's position may indicate the position of the top-leftmost sample of the block. Alternatively, the block's position may indicate the block's position within the frame.

[0989] For example, the block information may include block partitioning information, a reference region of the block, and a reference region available to the block. For example, coding parameters may include coding parameters described in embodiments, such as quantization parameters.

[0990] For example, objects associated with a block may include units and frames described in the embodiments. Units may include parameter sets, headers, bricks, slices, CTUs, CUs, PUs, and TUs. Units may include VPSs, SPSs, PPSs, APSs, and DPSs. Units may include picture headers, subpicture headers, slice headers, tile group headers, and tile headers. Frames may include previous frames, current frames, next frames, reference frames, etc.

[0991] One or more of the following exemplary [correction methods] may be used as a method for correcting an extrapolation filter derived in a specific region described in the embodiments.

[0992] [Correction Method 1]

[0993] For example, if an N-tap filter is inherited or derived, a method of changing K coefficients smaller than N into a filter can be applied, as in [Correction Method 1-1] to [Correction Method 1-5] below.

[0994] [Correction Method 1-1]

[0995] A method to remove coefficient(s) for sample(s) located relatively far from the current block can be applied. (In other words, cropping)

[0996] [Correction Method 1-2]

[0997] A method to remove coefficient(s) for sample(s) relatively close to the current block can be applied. (In other words, cropping)

[0998] [Correction Method 1-3]

[0999] A method to reduce the coefficients by using the average value of the coefficients at adjacent locations can be applied. (In other words, pooling)

[1000] [Correction Method 1-4]

[1001] A method to reduce the coefficients by using the maximum value among the coefficients at adjacent locations can be applied. (In other words, max pooling)

[1002] [Correction Method 1-5]

[1003] A method to reduce the coefficients by using the minimum value among the coefficients at adjacent locations can be applied. (In other words, min pooling)

[1004] In addition to [Correction Method 1-1] to [Correction Method 1-5], methods for deriving specific statistical values ​​described in the examples and filtering / sampling / scaling methods may be used to reduce the coefficients.

[1005] [Correction Method 2]

[1006] For example, if an N-tap filter is inherited or derived, the coefficient value can be adjusted in a way that derives the scaling adjustment value S.

[1007] For example, in the extrapolation filter-based prediction mode, if filtering is performed according to [Equation 5] below, the coefficient value can be corrected by the S value as in [Equation 6].

[1008] [Formula 5]

[1009] Pred(x, y) = ∑(C i × t(x - offsetX i , y - offsetY i )) + C N-1 × 2 bit depth - 1

[1010] [Equation 6]

[1011] Pred(x, y) = S × ∑(C i × t(x - offsetX i , y - offsetY i )) + C N-1 × 2 bit depth - 1

[1012] For example, in the extrapolation filter-based prediction mode, if filtering is performed according to [Equation 7] below, the coefficient value can be corrected by the S value as in [Equation 8].

[1013] [Equation 7]

[1014] Pred(x, y) = ∑(C i × t(x - offsetX i , y - offsetY i ))

[1015] [Equation 8]

[1016] Pred(x, y) = S × ∑(C i × t(x - offsetX i , y - offsetY i ))

[1017] Pred(x, y) = C N-1 (∑(C i × t(x - offsetX i , y - offsetY i )))

[1018] Pred(x, y) can be the predicted value at position (x, y) within the current block.

[1019] In [Equation 5] through [Equation 8], i can take one of the values ​​from 0 to N-2 in order by ∑. That is to say, ∑ can represent the sum of values ​​from 0 to N-2.

[1020] C i can be a filter coefficient. i can be a value greater than or equal to 0 and less than or equal to N-1.

[1021] t(x - offsetX i , y - offsetY i ) can be restoration samples or prediction samples.

[1022] Each of offsetX and offsetY can have an integer value greater than or equal to 1.

[1023] Here, for fixed candidate values, the candidate with the smallest template cost can be selected as S.

[1024] The template may consist of a reference region surrounding the current block that is the subject of the prediction. The template cost may be the difference between the predicted samples and the restored samples for the template.

[1025] Here, there may be P fixed candidate values. P can be an integer greater than or equal to 1. Each fixed candidate value can be one of a negative integer, 0, and a positive integer.

[1026]

[1027] Correction for predicted blocks using extrapolation filters

[1028] Below, an example is described of correcting the values ​​of samples within a predicted block based on an extrapolation filter according to the attributes / information of the current block.

[1029] In the embodiments, the predicted block may mean a predicted block, or a restoration block derived based on the predicted block.

[1030] Such correction may be used when the extrapolation filter of the luminance block or the extrapolation filter of the surrounding block of the chroma block is inherited and used by the merge technique; and when the extrapolation filter derived by the reference region surrounding the restored luminance block or chroma block is used; etc.

[1031] The corrections of the embodiments can be applied not only to the extrapolation filter of the chroma block but also to the corrections for samples within the block predicted by the extrapolation filter of the lumina block.

[1032] For example, the attributes of a block may include size, shape, position, and coding parameters. For example, the block's position may indicate the location of the block's top-to-leftmost sample. Alternatively, the block's position may indicate the block's location within the frame.

[1033] For example, the block information may include block partitioning information, a reference region of the block, and a reference region available to the block. For example, coding parameters may include coding parameters described in embodiments, such as quantization parameters.

[1034] For example, objects associated with a block may include units and frames described in the embodiments. Units may include parameter sets, headers, bricks, slices, CTUs, CUs, PUs, and TUs. Units may include VPSs, SPSs, PPSs, APSs, and DPSs. Units may include picture headers, subpicture headers, slice headers, tile group headers, and tile headers. Frames may include previous frames, current frames, next frames, reference frames, etc.

[1035] There is a possibility that discontinuities may occur at the boundary between the predicted block and the adjacent block using an extrapolation filter. In this case, Position Dependent Intra Prediction Combination (PDPC) can be performed on the predicted block.

[1036] For a prediction block based on an extrapolation filter for a chroma block, a correction utilizing the brightness difference between the samples of the corresponding lumina block (or, the surrounding reference samples of the lumina block) and the surrounding reference samples of the chroma block may be applied.

[1037] For example, a slope value (say, α) and a constant value (say, β or bias) between the values ​​of the surrounding reference samples of the corresponding luminance block and the surrounding reference samples of the chroma block can be derived, and the sample values ​​within the prediction block can be corrected using the derived slope value and the constant value.

[1038]

[1039] Entropy encoding / decoding of prediction information

[1040] In the embodiments, the encoding mode may mean a prediction mode. The encoding mode may represent a prediction mode using the extrapolation filter described in the embodiments. Alternatively, the encoding mode described in the embodiments may be one of the available prediction modes of the current blocks.

[1041] In the embodiments, the encoding information may refer to a prediction mode. The encoding information may represent prediction information for a prediction mode using the extrapolation filter described in the embodiments. Alternatively, the prediction information may include the encoding information described in the embodiments.

[1042] In the encoding device (110) and the decoding device (150), the encoding mode of the current chroma component block can be derived as described below.

[1043] The encoding mode may be determined based on information related to block partitioning. For example, the information related to block partitioning may include information used for partitioning blocks as described in the embodiments.

[1044] In one embodiment, when the lumina component block and the chroma component block have the same block partitioning structure (i.e., when a single tree type (i.e., SINGLE_TREE) is applied to the block), the encoding mode can be determined as described below.

[1045] For example, the prediction mode of a chroma component block can be set to be the same as the prediction mode of a corresponding lumina component block. For example, the prediction mode may be one of the prediction modes described in the embodiments, such as an intra prediction mode, an inter prediction mode, and an IBC mode.

[1046] In these and / or other embodiments, the intra prediction mode may mean at least one of the prediction modes described in the embodiments, such as: a non-directional prediction mode (e.g., DC mode, PLANAR mode, and block vector-based prediction); a directional prediction mode; a chroma component prediction mode; a subblock unit prediction mode; an intra partitioning prediction mode (e.g., Spatial Geometric Partitioning Mode (SGPM)); an intra prediction mode derivation using reference sample prediction (e.g., Template-based Intra Mode Derivation (TIMD)); an Intra Block Copy (IBC) mode; and an intra prediction mode to which an extrapolation filter is applied.

[1047] Additionally, if the corresponding luminance component block uses a prediction mode to which an extrapolation filter is applied, the residual block for the chroma component block is not encoded / decoded, and information regarding the residual block may not be signaled.

[1048] Here, indicator information indicating that information regarding the corresponding residual block is not signaled may not be encoded / decoded (entropy). In other words, it may be implicitly determined that the residual block is not encoded / decoded without being explicitly signaled. Information regarding the residual block may refer to information used to construct the residual block.

[1049] For example, the instruction information may include cu_cbf and tu_cbf, etc. tu_cbf may include at least one of tu_cbf_cb and tu_cbf_cr. That is to say, the instruction information may be information indicating whether information about a residual block is signaled and / or whether information about a residual block exists.

[1050] When the luminance component and the chroma component have the same block partitioning structure and the current chroma component block uses a prediction mode with an extrapolation filter applied (or, when the luminance component block corresponding to the current chroma component block uses a prediction mode with an extrapolation filter applied), the information required for encoding / decoding the current chroma component block can be derived from the encoding information of the luminance component block corresponding to the current chroma component block.

[1051] The required information can be derived from a specific Luma component block or a specific location of the Luma component block.

[1052] The required information may be derived from the encoding information of the luminance component block corresponding to a specific location described in the embodiments of the chroma component block. For example, the required information may be derived from the encoding information of the luminance component block corresponding to a sample location corresponding to the center of the chroma component block. For example, the required information may be derived from the encoding information of the luminance component block corresponding to the top-leftmost sample location of the chroma component block.

[1053] In one example, the encoding mode of the current chroma component block can be derived based on at least one of the current chroma component block; the current CTB; and the lumina component block or lumina component CTB corresponding to the chroma component block or chroma component CTB.

[1054] In one example, the encoding mode of the current chroma component block can be derived based on at least one of at least one encoding parameter among the current chroma component block; the current CTB; and the lumina component block or lumina component CTB corresponding to the chroma component block or chroma component CTB.

[1055] When the luma component and the chroma component have independent block partitioning structures (for example, when the tree type for a dual tree structure is DUAL_TREE_LUMA or DUAL_TREE_CHROMA), the encoding mode of the chroma component block can be determined from the encoding mode information of the chroma component block being encoded / decoded (entropy).

[1056] The encoding mode of the chroma component block may be one of the prediction modes described in the embodiments, such as the intra prediction mode, intra prediction mode, IBC mode, template matching prediction mode, and mode with an extrapolation filter applied, as well as the encoding mode of the luminance component block.

[1057] When the luminance component and the chroma component have mutually independent block partitioning structures and the current chroma component block uses a mode with an extrapolation filter applied, the information required for encoding / decoding the current chroma component block can be derived from the encoding information of the luminance component block corresponding to the current chroma component block. For example, the required information may include information used to configure the filter, such as filter coefficients, filter size, and filter shape, and may include information related to the filter described in the embodiments.

[1058] The required information can be derived from a specific Luma component block or a specific location of the Luma component block.

[1059] The required information may be derived from the encoding information of the luminance component block corresponding to a specific location described in the embodiments of the chroma component block. For example, the required information may be derived from the encoding information of the luminance component block corresponding to a sample location corresponding to the center of the chroma component block. For example, the required information may be derived from the encoding information of the luminance component block corresponding to the top-leftmost sample location of the chroma component block.

[1060] In one example, the encoding mode of a chroma component block can be derived based on at least one of the current chroma component block; the current CTB; and a lumina component block or lumina component CTB corresponding to the chroma component block or chroma component CTB.

[1061] In one example, the encoding mode of a chroma component block can be derived based on at least one of at least one encoding parameter among the current chroma component block; the current CTB; and the lumina component block or lumina component CTB corresponding to the chroma component block or chroma component CTB.

[1062]

[1063] Syntax elements

[1064] The following syntax elements may be used to carry out the embodiments. Predictive information may include the syntax elements described below. The names of the syntax elements may be illustrative merely for the sake of understanding and may be named as specific information or specific flags, etc.

[1065] EIPchroma_flag

[1066] EIPChroma_flag can indicate whether the current block is using a mode where an extrapolation filter is applied. The current block can be a chroma block.

[1067] The value of EIPChroma_flag can be explicitly signaled to encode / decode that the current chroma block uses a mode where an extrapolation filter is applied.

[1068] If the value of EIPChroma_flag is not explicitly signaled, the value of EIPChroma_flag can be determined by inference. In the above inference, information about the luminance block corresponding to the chroma block may be used.

[1069] For example, if the luminance block corresponding to the chroma block is not a block encoded / decoded using a mode where an extrapolation filter is applied, the value of EIPChroma_flag can be inferred to be 0.

[1070] For example, if the luminance block corresponding to the chroma block is a block encoded / decoded using a mode where an extrapolation filter is applied, the value of EIPChroma_flag can be inferred to be 1.

[1071] If the range of the reference region used to derive the filter is available around the current chroma block, the value of EIPChroma_flag can be explicitly signaled.

[1072] EIPchroma_merge_flag

[1073] EIPchroma_merge_flag can indicate whether filter coefficients from neighboring blocks encoded / decoded using the extrapolation filter mode are inherited by the current block when the current block uses the extrapolation filter mode.

[1074] If the current block uses an extrapolation filter mode, EIPchroma_merge_flag may indicate whether the chroma block inherits the filter coefficients of the corresponding luma block encoded / decoded in the extrapolation filter mode.

[1075] EIPChroma_merge_idx

[1076] EIPChroma_merge_idx may be the index value when the current block uses a mode with an extrapolation filter applied and the EIPChroma_merge_flag value exists.

[1077] In the embodiments, the existence of a specific value may mean that the specific value is explicitly signaled through the bitstream and / or that the specific value is derived according to a specific condition / rule. The non-existence of a specific value may mean that the specific value is not signaled through the bitstream and / or that the specific value is not defined as the specific value is not used.

[1078] If the current block is using a mode with an extrapolation filter applied and the EIPChroma_merge_flag value exists, EIPChroma_merge_idx can have an integer value less than or equal to max_eipChorma_merge_idx.

[1079] Max_eipchorma_merge_idx

[1080] Max_eipchorma_merge_idx can represent the maximum value of EIPChroma_merge_idx.

[1081] The value of Max_eipchorma_merge_idx can be an integer greater than or equal to 0.

[1082] The value of Max_eipchorma_merge_idx can be an integer of 8 or less.

[1083] If the value of Max_eipchorma_merge_idx is not signaled, the value of Max_eipchorma_merge_idx can be inferred as a predefined value.

[1084] EIPChroma_filter_refine_flag

[1085] EIPChroma_filter_refine_flag may indicate whether the current block uses a mode where an extrapolation filter is applied, and whether a method to correct the filter values ​​is applied to the above filter when using an inherited filter.

[1086] Here, the inherited filter may include a filter applied to a corresponding luma block; and an extrapolation filter applied within a restored chroma component region, and other filters described as objects of inheritance in the embodiments.

[1087] For example, if the value of EIPChroma_merge_flag exists, the value of EIPChroma_filter_refine_flag can be inferred as 1.

[1088] EIPChroma_predictor_refine_flag

[1089] EIPChroma_predictor_refine_flag may indicate whether a method to correct the values ​​of samples is applied to samples within the filtered prediction block when the current block uses a mode where an extrapolation filter is applied.

[1090] EIPChroma_MM_flag

[1091] EIPChroma_MM_flag can indicate whether a method using two or more filters is applied when the current block uses a mode where extrapolation filters are applied.

[1092] EIPChroma_Idx

[1093] EIPChroma_Idx may be an index indicating which of the available filters that can be derived is used when the current block uses a mode where an extrapolation filter is applied and not a merge mode.

[1094] For example, the total number of available filters that can be derived may be determined by the size of the current block and a surrounding reference sample area that can be referenced. Additionally, the method for determining the number of filters described in the embodiments may be used to determine the total number of available filters that can be derived.

[1095]

[1096] Entropy encoding and entropy decoding of prediction information

[1097] As described above, the prediction information and / or the encoded information may be entropy encoded in the encoding device (110). The entropy-encoded prediction information and / or the entropy-encoded encoded information may be entropy-decoded in the decoding device (150).

[1098] At least one of the syntax elements described in the embodiments can be entropy encoded / decoded using at least one of the binarization; debinarization; and entropy encoding / decoding methods described in the embodiments.

[1099] For example, the binarization; debinarization; and entropy encoding / decoding methods described in the embodiments may include se(v), sek(v), ue(v), uek(v), f(n), tu(v), tb(v), ae(v), b(8), i(n), u(n), and unary binarization / debinarization methods.

[1100] When encoding information related to a prediction mode based on an extrapolation filter is encoded / decoded, a context model for the encoding information can be determined using specific information.

[1101] For example, specific information may include encoding information related to a prediction mode based on an extrapolation filter of surrounding blocks; encoding information related to a prediction mode based on a previously encoded / decoded extrapolation filter; and / or information related to attributes of the current unit / block. For example, information related to attributes of the current unit / block may include information regarding the depth of the current unit / block and information regarding the size of the current unit / block.

[1102] When encoding information related to a prediction mode based on an extrapolation filter is encoded / decoded, specific information can be used as a prediction value for the above encoding information.

[1103] For example, specific information may include encoding information related to a prediction mode based on an extrapolation filter of surrounding blocks; encoding information related to a prediction mode based on a previously encoded / decoded extrapolation filter; and / or information related to attributes of the current unit / block. For example, information related to attributes of the current unit / block may include information regarding the depth of the current unit / block and information regarding the size of the current unit / block.

[1104] The encoding / decoding process of the current block may not be limited to any single embodiment. At least a portion of a specific embodiment may be applied to the encoding / decoding process of the current block. Additionally, a combination of multiple embodiments may be applied to the encoding / decoding process of the current block. Here, the combination of multiple embodiments may include a portion of each embodiment.

[1105]

[1106] Restoration region of the extrapolation filter

[1107] FIGS. 16a to 16c show the restoration regions for an extrapolation filter according to one example.

[1108] In FIGS. 16a to 16c, three types of restoration regions for the extrapolation filter are illustrated.

[1109] As illustrated in FIGS. 16a through 16c, the restoration regions for the extrapolation filter can be specified by coding parameters such as leftSize and aboveSize. leftSize can be used to specify the width of the restoration region. aboveSize can be used to specify the height of the restoration region.

[1110] As exemplified in FIG. 16a, the restoration area may include a rectangular area diagonally adjacent to the top-left corner of the current block; a rectangular area adjacent to the top of the current block; and a rectangular area adjacent to the left of the current block. That is to say, the restoration area of ​​the extrapolation filter may be part of an already restored area to the left or top of the current block.

[1111] As exemplified in FIG. 16b, it may include a rectangular area adjacent to the top of the current block. That is to say, the restored area of ​​the extrapolation filter may be part of the already restored area at the top of the current block.

[1112] As exemplified in FIG. 16c, it may include a rectangular area adjacent to the left of the current block. That is to say, the restored area of ​​the extrapolation filter may be part of the already restored area to the left of the current block.

[1113] Within the restoration region, an extrapolation filter with a size of fWidth × fHeight can be used.

[1114]

[1115] Composition of the list for the extrapolation filter

[1116] If an extrapolation filter is applied to the current block, a merge flag such as EIPchroma_merge_flag may be signaled.

[1117] The merge flag can indicate whether the extrapolation filter of the current block is inherited from a block encoded / decoded using a mode in which the extrapolation filter was previously applied.

[1118] If the value of the merge flag is true, a merge list for an extrapolation filter can be constructed using the candidates described in the embodiments, such as block vectors of spatially adjacent blocks (so-called, spatial candidate); block vectors of spatially non-adjacent blocks (so-called, non-adjacent candidate); temporal candidate; and history-based candidate.

[1119] The positions of these candidates within the merge list and the order in which they are included in the merge list may be the same as the positions and order described in other merge candidate lists of the embodiments.

[1120] Additionally, a merge index, such as EIPChroma_merge_idx, indicating which candidate in the merge list has been selected may be additionally signaled. Among the candidates in the merge list, the candidate indicated by the merge index may be selected.

[1121] The filter type and filter coefficients of the selected candidates in the merge list can be inherited as the extrapolation filter of the current block and can be used for encoding / decoding the current block.

[1122] If the value of the merge flag is false, the extrapolation filter can be derived from the surrounding recovery samples of the current block. Additionally, related syntax elements may be signaled to indicate which of the three types of recovery regions is used for the current block and which of the filter forms described in the embodiments is used for the current block.

[1123] The selected filter can construct an auto-correlation matrix and a cross-correlation vector by moving horizontally or vertically at 1-pixel intervals within the selected restoration area.

[1124] Similar to a Cross-Component Linear Model (CCCM) with L-2 regularization, coefficients can be calculated from the autocorrelation matrix and the cross-correlation vector. The coefficients can be constructed according to [Equation 9] below.

[1125] [Formula 9]

[1126]

[1127] λ can be a regulation parameter. L2-normalization can be achieved by adding a diagonal matrix λI to the 'ATA' matrix.

[1128] Some of the coefficients can be relaxed (that is to say, unregularized). In this case, when the i-th coefficient is relaxed, the entry (i-th row, i-th column) of the diagonal matrix λI can be set to 0.

[1129] In the scheme of a normalized extrapolation filter, it can be set as shown in [Equation 10] below.

[1130] [Formula 10]

[1131] λ = M - p EIP

[1132] Here, p EIP = 15 can be the number of filter tabs within the extrapolation filter.

[1133] When the number of input samples nSamples is less than or equal to 2024, M can be 192. Also, therefore, λ can be 2880.

[1134] Otherwise, M can be 128. Also, therefore, λ can be 1920.

[1135] The bias term of the extrapolation filter can be relaxed (so to speak, denormalized). The bottom-left entry of the diagonal matrix can be set to 0.

[1136] In one example, after a prediction sample of the current block is generated using an extrapolation filter, an intra prediction mode can be induced by applying a DIMD process to the prediction sample.

[1137] Specifically, for each prediction sample, horizontal and vertical gradients can be calculated, and a gradient histogram can be generated using these gradients. Subsequently, an LFNST, NSPT, or MTS transformation set can be determined using an intra-prediction mode corresponding to the largest histogram count.

[1138]

[1139] Extrapolation filter-based intra prediction guided by block vectors

[1140] Figure 17 shows a reference area for the current block and extrapolation filter according to one example.

[1141] In intra prediction based on an extrapolation filter guided by a block vector, instead of directly using the spatial reconstruction region surrounding the current block, the block vector may be used to determine a reference region for obtaining information to construct the parameters of the extrapolation filter.

[1142] An extrapolation filter-based intra prediction mode guided by a block vector may be a submode of the extrapolation filter-based intra prediction mode.

[1143] When the current block is encoded / decoded using a submode of extrapolation filter-based intra-prediction guided by a block vector, extrapolation filter parameters calculated by the method of the example; and surrounding reconstructed samples; can be used to generate the prediction block.

[1144] For intra prediction based on an extrapolation filter guided by a block vector, the shape of the restoration region may be limited to a specific restoration region type, rather than the restoration regions for the extrapolation filter described in the embodiments, such as those described with reference to FIGS. 16a to 16c.

[1145] For example, a square restoration region may be allowed for an extrapolation filter-based intra-prediction mode guided by a block vector.

[1146] In Fig. 17, the reference region of an extrapolation filter-based intra-prediction guided by block vectors is shown.

[1147] A flag may be signaled to indicate whether to use an extrapolation filter-based intra-prediction mode guided by a block vector.

[1148] The block vector of the embodiment can be determined by a search and derivation method for the block vector described in the embodiment, such as Intra Template Matching Prediction (IntraTMP).

[1149]

[1150] Multi-model extrapolation filter

[1151] In the configuration of the extrapolation filters of the embodiments, a Multi-Model (MM) mode may be used. The MM mode may be a mode in which the classification according to the description below is additionally applied to the extrapolation filter-based intra prediction mode described in the embodiments.

[1152] In MM mode, EIP training samples can be classified into two groups, and two individual extrapolation filters can be derived by using the samples from each of the two groups as the extrapolation filter training set.

[1153] The classification process can be based on the reference values ​​and thresholds of the extrapolation filter training samples.

[1154] The threshold can be set to the average value of all samples within the restoration area surrounding the current block.

[1155] The reference value of the current extrapolation filter training sample can be calculated by one or two input samples.

[1156] Extrapolation filter reference samples having reference values ​​below a threshold can be classified into a first group. Extrapolation filter reference samples having reference values ​​greater than the threshold can be classified into a second group.

[1157] Two extrapolation filters can be constructed according to [Equation 11] below.

[1158] [Equation 11]

[1159]

[1160] When deriving the coefficients of the extrapolation filter in MM mode, the regularization parameter λ can be adaptively selected based on the number of training samples.

[1161] The same classification rules can be applied to the prediction of the current block. Prediction samples can be calculated in the same way based on the reference values ​​and thresholds of the current inputs.

[1162] When a prediction using an extrapolation filter is performed for the current block and a merge mode using an extrapolation filter is not performed, an MM mode flag indicating whether MM mode is used may be signaled.

[1163] In MM mode, the number of derived filter candidates can be adjusted so that the total number of derived filter candidates (i.e., all of the candidates of the extrapolation filter-based prediction mode of the example; and the candidates of the MM mode;) does not exceed the previous number.

[1164] The maximum template size of the extrapolation filter derivation process in MM mode can be set to 8.

[1165]

[1166] Improvement of prediction mode based on extrapolation filter

[1167] The prediction mode based on the extrapolation filter described in the embodiments may be fused with one or more other intra-prediction modes described in the embodiments. Such fusion is described below.

[1168] Other intra prediction modes may include non-directional prediction modes (e.g., DC mode, PLANAR mode, and block vector-based prediction); directional prediction modes; chroma component prediction modes; subblock unit prediction modes; intra partitioning prediction modes (e.g., Spatial Geometric Partitioning Mode (SGPM)); intra prediction mode derivation using reference sample prediction (e.g., Template-based Intra Mode Derivation (TIMD)); and Intra Block Copy (IBC) modes; etc.

[1169] Additionally, other intra-prediction modes may include modes that include / use the intra-prediction described in the embodiments.

[1170] A final prediction block can be generated by weighted summing a prediction block predicted using a different intra-prediction mode; and a block predicted by a method based on an extrapolation filter. Alternatively, a final prediction block can be generated by weighted summing a prediction block predicted using a different intra-prediction mode; and a block predicted by a method based on an extrapolation filter.

[1171] For example, an average fusion of a prediction block generated using Planar mode and a prediction block generated based on an extrapolation filter can be generated, and the average fusion can be used as a final prediction block. Alternatively, a final prediction block can be generated based on this average fusion.

[1172] For example, an average fusion of a prediction block generated using IntraTMP mode and a prediction block generated based on an extrapolation filter can be generated, and the average fusion can be used as a final prediction block. Alternatively, a final prediction block can be generated based on this average fusion.

[1173] An extrapolation filter for the current prediction block can be derived using blocks predicted by other intra prediction modes.

[1174] For example, an extrapolation filter can be derived by utilizing samples within a predicted block using Planar mode. The predicted block to which this extrapolation filter is applied can be used as the final predicted block.

[1175] For example, an extrapolation filter can be derived by utilizing samples within a predicted block using IntraTMP mode. The prediction block to which this extrapolation filter is applied can be used as the final prediction block.

[1176] For example, an extrapolation filter can be derived using samples within reference blocks corresponding to multiple block vectors. A prediction block to which this extrapolation filter is applied can be used as a final prediction block.

[1177] In performing predictions based on the extrapolation filters of the embodiments, a prediction combined with one or more inter-predictions may be performed.

[1178] The combination of a specific prediction and another prediction may refer to a combination between a prediction block generated by a specific prediction and a prediction block generated by another prediction. The combination may be a weighted combination. In other words, a weighted combination may refer to a weighted sum.

[1179] For example, a first prediction sample can be derived by performing an inter-prediction on the current block. A second prediction sample can be derived using a prediction method based on an extrapolation filter. A third prediction sample can be derived by the weighted sum of the first prediction sample and the second prediction sample. The third prediction sample may be a prediction sample of the final prediction block for the current block.

[1180] Here, the inter prediction of the inter prediction may be at least one of the inter prediction modes and modes using inter prediction described in embodiments such as merge mode; Advanced Motion Vector Prediction (AMVP) mode; Combined Inter Intra Prediction (CIIP) mode; and Local Illumination Compensation (LIC) mode.

[1181]

[1182] Intra prediction mode based on block vectors

[1183] In the embodiments described below, the configuration of block vector candidates and final block vectors for a prediction mode using block vector information; a prediction mode using IBC; etc. is described.

[1184] Block vector information may mean a block vector as described in the embodiments; information representing the block vector; and information used to derive the block vector.

[1185] Prediction modes using block vector information may include Intra Template Matching Prediction (IntraTMP) mode; Intra Block Copy (IBC) mode; and a method of fusing one or more predictors derived by the block vector with predictors of other prediction modes. Samples within the reference block pointed to by the block vector may be used as predictors.

[1186] In the process of encoding / decoding an image, a picture may be encoded / decoded using one of the prediction modes described in the embodiments, such as intra prediction, inter prediction, template matching prediction, and IBC mode.

[1187] Additionally, during the encoding / decoding process of the image, a block may be encoded / decoded using one of the prediction modes described in the embodiments, such as intra prediction, inter prediction, template matching prediction, and IBC mode.

[1188] In addition, a template matching prediction mode using template matching prediction may be included as one of the prediction modes used in the encoding / decoding process of such images / blocks.

[1189]

[1190] Prediction using intra-block copy mode

[1191] Figure 18 illustrates the concept of an intra-block copy mode according to one example.

[1192] In intra-block copy mode, the area most similar to the current block can be searched among the restoration areas surrounding the current block. The area corresponding to the current block determined by the search can be used as the reference block for the current block.

[1193] In Fig. 18, the current block and reference block in intra-block copy mode are shown.

[1194] In FIG. 18, an example is given where the sample position of the top-leftmost sample of the current block is (x0, y0), the horizontal length X is a negative integer, the vertical length difference Y is a negative integer, and the sample position of the top-leftmost sample of the reference block is (x0+ X, y0+ Y).

[1195] In the embodiments, the horizontal length for the objects may mean the difference between the x-coordinate positions of the objects. The vertical length for the objects may mean the difference between the y-coordinate positions of the objects.

[1196] The horizontal length X may be the distance from the x-coordinate position of the top-leftmost sample of the current block to the x-coordinate position of the top-leftmost sample of the reference block. The vertical length Y may be the distance from the y-coordinate position of the top-leftmost sample of the current block to the y-coordinate position of the top-leftmost sample of the reference block.

[1197] A block vector can be a vector (X, Y). That is to say, a block vector can be a vector from the current block to the reference block. A block vector can be regarded similarly to the motion vector of an inter-prediction. For example, the description of the motion vector of the inter-prediction of the embodiments may also apply to the block vector. However, in an intra-block copy using a block vector, only the block within the current picture containing the current block may be specified as the reference block.

[1198] Here, the reference block of the current block may refer to the prediction block for the current block. In other words, the sample values ​​of the reference block can be used as the sample values ​​of the prediction block. Alternatively, the prediction block can be derived based on the reference block.

[1199] The block vector derived by template matching prediction can be a block vector of intra-block copy mode.

[1200]

[1201] Prediction using template matching

[1202] Figure 19 illustrates the concept of template matching prediction according to one example.

[1203] In template matching prediction, a portion of the restoration regions surrounding the current block may be used as a matching template. Within the search region, the region most similar to the current template may be searched. The searched most similar region may be determined as the matching template. Here, the most similar region may be an area that has the same shape as the current template and has the minimum error cost compared to the current template. The search region may be the entire restoration region or a portion of a single picture.

[1204] In other words, unlike intra-block copying where the reference block is determined based on the similarity between the current block and the corresponding block, template matching prediction may be a prediction where the matching template is determined based on the similarity between the current template of the current block and the corresponding template of the corresponding block, and the reference block is determined by the matching template.

[1205] When a matching template is determined by search, a reference block corresponding to the current block can be determined. For example, the relative difference between the location of the current template and the location of the current block may be equal to the relative difference between the location of the matching template and the location of the reference block. In other words, the location of the reference block can be derived by applying the relative difference between the location of the current template and the location of the current block to the location of the matching template.

[1206] In FIG. 19, an example is given where the sample position of the top-leftmost sample of the current block is (x0, y0), the horizontal length X is a negative integer, the vertical length Y is a negative integer, and the sample position of the top-leftmost sample of the reference block is (x0+ X, y0+ Y).

[1207] The horizontal length X may be the distance from the x-coordinate position of the top-leftmost sample of the current block to the x-coordinate position of the top-leftmost sample of the reference block. Alternatively, the horizontal length X may be the distance from the x-coordinate position of the current template to the x-coordinate position of the reference template. Alternatively, the horizontal length X may be the distance from the x-coordinate position of the top-leftmost sample of the current template to the x-coordinate position of the top-leftmost sample of the reference template.

[1208] The vertical length Y may be the distance from the y-coordinate position of the top-leftmost sample of the current block to the y-coordinate position of the top-leftmost sample of the reference block. Alternatively, the vertical length Y may be the distance from the y-coordinate position of the current template to the y-coordinate position of the reference template. Alternatively, the vertical length Y may be the distance from the y-coordinate position of the top-leftmost sample of the current template to the y-coordinate position of the top-leftmost sample of the reference template.

[1209] A block vector can be a vector (X, Y). That is to say, a block vector can be a vector from the current block to the reference block. Alternatively, a block vector can be a vector from the current template to the matching template. A block vector can be regarded similarly to the motion vector of inter-prediction. For example, the description of the motion vector of inter-prediction in the embodiments may also apply to block vectors. However, in an intra-block copy using a block vector, only the block within the current picture containing the current block may be specified as the reference block.

[1210] The block vector derived by this template matching prediction can be a template matching block vector.

[1211]

[1212] Construction of intra prediction candidates based on block vectors

[1213] Below, a block vector-based intra-prediction technique and a method for using the intra-prediction mode derived by the intra-prediction technique are described.

[1214] Block vector-based intra prediction technology may include an Intra Template Matching Prediction (IntraTMP) mode and an Intra Block Copy (IBC) mode.

[1215] In block vector-based intra-prediction technology, multiple candidates can be configured, and a final prediction block can be generated.

[1216] Here, the number of multiple candidates may be a non-zero integer N. One or more selected candidates from the total number of multiple candidates may be used to generate a prediction block.

[1217] Candidates configured for the prediction of the current block using block vectors can be called block vector candidates.

[1218] Candidates can be arranged in ascending or descending order of error cost.

[1219] For example, the error cost can be constructed using one or more of the Sum of Absolute Differences (SAD), Sum of Absolute Transformed Differences (SATD), and Mean Removal-SAD (MR-SAD).

[1220] For example, 20 block vector candidates can be constructed using the SAD error cost.

[1221] The composition of candidates can be performed by a search process in a defined area or other areas.

[1222] Here, the defined area may be an area determined by the size of the current block or a fixed area.

[1223] During the process of constructing candidates, block vector candidates outside the defined region can become merge candidates.

[1224] Here, if the restored blocks surrounding the current block are encoded / decoded using a block vector-based intra-prediction technique, the first block vector of the above restored blocks may be included as a merge candidate for the current block.

[1225] Here, if the block indicated by the above block vector is also encoded / decoded using a block vector-based intra-prediction mode, a block vector shifted from the first block vector by the second block vector of the above indicated block may be included as a merge candidate for the current block.

[1226] Candidates configured according to this method can be called Auto Relocated-Block Vector Predictors (AR-BVP).

[1227] In one embodiment, in a method for constructing candidates for a block vector-based intra-prediction technique, a plurality of block vector candidates can be constructed through a method that can be used when a reference picture is available.

[1228] Here, whether a reference picture is available can be determined based on the slice information of the current block.

[1229] For example, if the slice containing the current block is a P slice or a B slice, a non-zero integer number of reference picture(s) may be used.

[1230] A candidate constructed through a reference picture can be called a temporal candidate.

[1231] A block vector-based intra-prediction technique for a current block may be configured to include one or more of: a block vector of a spatially adjacent block (i.e., a spatial candidate); a block vector of a spatially non-adjacent block (i.e., a non-adjacent candidate); a temporal candidate; a merge candidate; a candidate generated based on other candidates described in the embodiments; and a candidate generated by combining a plurality of candidates.

[1232] When a temporal candidate for the current block is constructed, the collated block and the adjacent block of the collated block can be used to derive the block vector.

[1233] For example, if a juxtaposed block is encoded / decoded using a block vector-based intra-prediction technique, the block vector of the juxtaposed block may be included as a candidate for the block vector of the current block.

[1234] Here, if the block vector index of the reference block is an integer M greater than 0, the candidate corresponding to M and the previous candidate corresponding to M may be included as block vector candidates of the current block.

[1235] For example, if the block vector index of the reference block is 4, candidates from index 0 to index 4 of the reference block may be included as block vector candidates of the current block.

[1236] For example, if adjacent blocks surrounding a juxtaposed block are encoded / decoded using a block vector-based intra-prediction technique, the block vectors of these adjacent blocks may be included as candidates for the block vector of the current block.

[1237] Here, the number of adjacent blocks being referenced can be a non-zero integer, such as 4, for example.

[1238] In constructing candidates for block vector-based intra-prediction technology, multiple block vector candidates can be constructed through the combination of block vectors.

[1239] Block vector-based intra-prediction technology can construct block vector candidates using candidates constructed by exploration; and merge candidates.

[1240] A merge candidate may refer to a candidate derived from an area outside the current block's search area.

[1241] The number of block vector candidates for the current block can be a non-zero integer N.

[1242] Here, N block vector candidates may simultaneously include merge candidates; and candidates configured by search; or may include only one of merge candidates; and candidates configured by search.

[1243] In one embodiment, the merge candidate may be configured to include one or more of: a block vector of spatially adjacent blocks (i.e., a spatial candidate); a block vector of spatially non-adjacent blocks (i.e., a non-adjacent candidate); a temporal candidate; an Auto Relocated-Block Vector Predictor (AR-BVP); a history-based Block Vector Predictor (i.e., a history-based Block Vector Predictor (HBVP)); a candidate derived by the average combination of multiple candidates (i.e., a Pairwise Average Block Vector Predictor (PABVP); and a candidate derived by the weighted combination of multiple candidates.

[1244] A candidate derived by combining these multiple candidates can be ...

Claims

1. A step for inducing a prediction mode for the current block; A step of constructing a reference sample for prediction regarding the current block above; and Step of performing the prediction for the current block using the above reference sample and the above derived prediction mode A video decoding method including 2. In Paragraph 1, An image decoding method in which the above prediction is performed using an intra-prediction mode in which an extrapolation filter is applied to the above current block.

3. In Paragraph 2, An image decoding method in which, in an intra-prediction mode in which the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block is applied to the chroma block.

4. In Paragraph 2, An image decoding method in which correction is performed on a prediction block to which the extrapolation filter is applied to the current block.

5. In Paragraph 2, An image decoding method in which the extrapolation filter for the current block above is inherited from the extrapolation filter of the restored block.

6. In Paragraph 2, An image decoding method in which the extrapolation filter of the current block is derived using samples within the restored region surrounding the current block.

7. In Paragraph 2, An image decoding method in which correction is performed on the coefficient values ​​of the extrapolation filter of the current block.

8. A step for deriving a prediction mode for the current block; A step of constructing a reference sample for prediction regarding the current block above; and Step of performing the prediction for the current block using the above reference sample and the above derived prediction mode A video encoding method including 9. In Paragraph 8, An image encoding method in which the above prediction is performed using an intra-prediction mode in which an extrapolation filter is applied to the above current block.

10. In Paragraph 9, An image encoding method in which, in an intra-prediction mode in which the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block is applied to the chroma block.

11. In Paragraph 9, An image encoding method in which correction is performed on a prediction block to which the extrapolation filter is applied to the current block.

12. In Paragraph 9, An image encoding method in which the extrapolation filter for the current block above is inherited from the extrapolation filter of the restored block.

13. In Paragraph 9, An image encoding method in which the extrapolation filter of the current block is derived using samples within a restored region surrounding the current block.

14. In Paragraph 9, An image encoding method in which correction is performed on the coefficient values ​​of the extrapolation filter of the current block.

15. A computer-readable recording medium storing a bitstream generated by the image encoding method of claim 8.

16. In a computer-readable recording medium storing a bitstream, the bitstream is, Predictive information Includes, A prediction mode for the current block is derived using the above prediction information, and A reference sample for prediction regarding the above current block is configured, and A computer-readable recording medium that performs the prediction for the current block using the above reference sample and the above-derived prediction mode.

17. In Paragraph 16, A computer-readable recording medium in which the above prediction is performed using an intra-prediction mode in which an extrapolation filter is applied to the above current block.

18. In Paragraph 17, A computer-readable recording medium in which, in an intra-prediction mode in which the extrapolation filter for the current block is applied, the extrapolation filter derived for the luminance block is applied to the chroma block.

19. In Paragraph 17, A computer-readable recording medium in which correction is performed on a prediction block to which the extrapolation filter is applied to the current block.

20. In Paragraph 16, A computer-readable recording medium in which the extrapolation filter for the current block above is inherited from the extrapolation filter of the restored block.

Images