Image encoding / decoding method and device, and recording medium

Abstract

An image encoding / decoding method and device for performing prediction, and a recording medium are disclosed. The encoding / decoding method comprises the steps of: deriving a prediction mode for a target block from among various possible prediction modes; using selection, filtering, sampling, interpolation and the like for a reference sample so as to configure a reference sample for prediction of the target block; and using the reference sample and the derived prediction mode so as to perform prediction on the target block. The prediction mode can be an intra prediction mode, and the derived intra prediction mode is stored for the target block.

Description

Method, device and recording medium for video encoding / decoding

[0001] The present invention relates to a method, apparatus, and recording medium for image encoding / decoding. Specifically, embodiments of the present invention disclose a method, apparatus, and recording medium for image encoding / decoding, which are related to performing prediction according to an induced prediction mode.

[0002] This invention claims the benefit of the filing dates of Korean Patent Application No. 10-2024-0048186, filed April 9, 2024, Korean Patent Application No. 10-2024-0089204, filed July 5, 2024, and Korean Patent Application No. 10-2025-0046251, filed April 9, 2025, the entire contents of which are incorporated herein by reference.

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

[0004] Users demand higher resolution and higher quality video. To meet these demands, video encoding / decoding technologies tailored to these needs are required. Video encoding technology can create compressed video by compressing the video representing the images into a smaller amount of data. Video decoding technology can use the compressed video to create reconstructed images.

[0005] When it comes to video encoding / decoding, various technologies exist, including segmentation, prediction, transformation, quantization, filtering, and entropy encoding / decoding. By introducing, modifying, improving, and combining these diverse technologies, video and images can be compressed, transmitted, and stored more effectively.

[0006] One embodiment may provide a device, method and recording medium for performing prediction according to an induced prediction mode.

[0007] A method for decoding an image is provided, comprising: a step of deriving a prediction mode for a target block on one side; a step of constructing a reference sample for prediction for the target block; and a step of performing the prediction for the target block using the reference sample and the derived prediction mode.

[0008] A candidate list for the above prediction of the above target block can be constructed.

[0009] The above candidate list can be derived using a Histogram of Gradients (HoG).

[0010] The above HoG can be configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

[0011] The above HoG can be configured based on the amount of change in pixel values ​​of predicted pixels of the predicted block.

[0012] An intra prediction mode derived using a restoration block or block vector of the target block can be stored as an intra prediction mode of the target block.

[0013] The storage of the derived intra prediction mode can be performed when the prediction does not use an intra prediction directional mode.

[0014] The above prediction can derive an intra prediction mode using analysis of reference samples.

[0015] In another aspect, a video encoding method is provided, comprising: a step of deriving a prediction mode for a target block; a step of constructing a reference sample for prediction for the target block; and a step of performing the prediction for the target block using the reference sample and the derived prediction mode.

[0016] A candidate list for the above prediction of the above target block can be constructed.

[0017] The above candidate list can be derived using a Histogram of Gradients (HoG).

[0018] The above HoG can be configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

[0019] The above HoG can be configured based on the amount of change in pixel values ​​of predicted pixels of the predicted block.

[0020] An intra prediction mode derived using a restoration block or block vector of the target block can be stored as an intra prediction mode of the target block.

[0021] The storage of the derived intra prediction mode can be performed when the prediction does not use an intra prediction directional mode.

[0022] The above prediction can derive an intra prediction mode using analysis of reference samples.

[0023] A computer-readable recording medium for storing a bitstream generated by the above image encoding method may be provided.

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

[0025] A candidate list for the above prediction of the above target block can be constructed.

[0026] The above candidate list can be derived using a Histogram of Gradients (HoG).

[0027] The above HoG can be configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

[0028] The above HoG can be configured based on the amount of change in pixel values ​​of predicted pixels of the predicted block.

[0029] An intra prediction mode derived using a restoration block or block vector of the target block can be stored as an intra prediction mode of the target block.

[0030] A device, method and recording medium for performing prediction according to an induced prediction mode are provided.

[0031] Figure 1 illustrates a system for video coding according to one embodiment.

[0032] Figure 2 shows a segmentation structure of an image according to one embodiment.

[0033] Figure 3 illustrates the structure of intra prediction according to one embodiment.

[0034] Figure 4 shows the structure of inter prediction to explain the inter prediction process according to one embodiment.

[0035] Figure 5 shows the order in which spatial candidates are added to the candidate list according to one embodiment.

[0036] Figure 6 illustrates multiple in-loop filters according to an example.

[0037] Figure 7 shows the structure of entropy encoding and entropy decoding according to an example.

[0038] FIG. 8 is a flowchart of a method for predicting a target block and a method for generating a bitstream according to one embodiment.

[0039] FIG. 9 is a flowchart of a method for predicting a target block using a bitstream according to one embodiment.

[0040] Figure 10 is a flowchart of a prediction method according to an example.

[0041] Figure 11 shows neighboring reconstructed samples used for DIMD chroma mode according to an example.

[0042] Figure 12 shows a search area for intra template matching according to an example.

[0043] Figure 13 illustrates the use of IntraTMP block vectors for IBC blocks.

[0044] Figure 15 shows the area of ​​samples used for analysis within a target block according to an example.

[0045] Figure 16 shows the area of ​​samples used for analysis determined according to the size of the target block according to an example.

[0046] Figure 17 shows the locations of samples within a reference block according to an example.

[0047] Figure 18 illustrates a mapping of information using the location of a target block and the block vector of an adjacent block according to an example.

[0048] Figure 19 illustrates a mapping of information using the location of adjacent blocks and the block vector of the adjacent blocks according to an example.

[0049] Figure 20 shows the surrounding blocks of another target block as an example.

[0050] Figure 21 illustrates adjacent blocks and non-adjacent blocks of a target block according to an example.

[0051] The present invention is capable of various modifications. Furthermore, the present invention may have various embodiments. Specific embodiments are described in detail in the drawings and detailed description.

[0052] It should be understood that the specific examples are not intended to limit the invention to specific embodiments, and that all modifications, equivalents, and substitutes falling within the spirit and scope of the invention are intended to be encompassed within the scope of the invention.

[0053] The embodiments are described in sufficient detail to enable those skilled in the art to practice them. It should be understood that the various embodiments, while different from each other, are not necessarily mutually exclusive. For example, it should be understood that the shapes, structures, and characteristics described in connection with one embodiment may be applied to or implemented in other embodiments without departing from the spirit and scope of the present invention. Furthermore, it should be understood that the positions or arrangements of components within one embodiment may be modified without departing from the spirit and scope of the present invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the exemplary embodiments, if properly described, is defined only by the appended claims and all equivalents to those claimed by such claims.

[0054] A detailed description of the embodiments described below may refer to the drawings for the embodiments. Any description described in the drawings or the descriptions shown in the drawings may be considered part of the detailed description. In the drawings, similar reference numerals may designate the same or similar functions throughout various aspects. The dependencies between components may not be limited to those depicted in the drawings.

[0055] In the embodiments, a singular expression may include, and may be limited to, and / or restricted by, a plural expression, unless the context clearly excludes a plural expression. That is, expressions such as “at least one” and “one or more” in the embodiments may be replaced with “plural.” Terms such as “ / ,” “and / or,” “at least one of,” and “one or more of” described for a plurality of items may mean 1) one item of the plurality of items, 2) some of the plurality of items, 3) a combination of some of the plurality of items, or 4) a combination of the plurality of items. Furthermore, a plural expression may be replaced with a singular expression. The plural may mean an integer greater than or equal to 1, 2, 3, 4, or 5.

[0056] In the embodiments, terms related to numbers, such as "first" and "second," may be used to describe various components. These terms are used only to distinguish one component from another and do not limit the components. For example, without departing from the scope of the present invention, the first component could be referred to as the second component, and similarly, the second component could also be referred to as the first component.

[0057] When a first component transmits (or provides) information to a second component, it can mean that the first component directly transmits information to the second component, or it can mean that the first component transmits information to the second component via another third component. Here, the information that the second component receives (or obtains) can be information transmitted by the first component, or information generated by applying a specific process to information transmitted by the first component.

[0058] The components of the embodiments may be depicted independently to represent different characteristic functions, and this does not imply that each component corresponds to a separate hardware or software configuration unit. That is, the components of the embodiments may be distinguished and listed 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 aforementioned component. Embodiments in which such components are integrated and embodiments in which such components are separated are also included in the scope of the present invention, as long as they do not depart from the essence of the present invention.

[0059] The terms used in the embodiments are used only to describe specific embodiments and are not intended to limit the present invention. In the embodiments, terms such as "comprise" or "have" indicate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the embodiments. These terms do not preclude the presence or addition of other features, numbers, steps, operations, components, parts, or combinations thereof that are not explicitly described in the embodiments. In other words, the description of "comprising" a specific component of an embodiment does not exclude other components other than the specific component, and means that additional components may also be included in the scope of the embodiments of the present invention or the technical idea of ​​the present invention.

[0060] Some of the components of the embodiments may be optional, not essential components for performing the essential functions of the present invention. These optional components may be used to improve performance. The embodiments may be implemented as a structure that includes only the essential components required to implement the essence of the embodiments, excluding the optional components. Such a structure is also within the scope of the embodiments.

[0061] Hereinafter, embodiments are described in detail with reference to the attached drawings to enable those skilled in the art to easily implement the embodiments. In describing the embodiments, if a detailed description of a related known configuration or function is judged to obscure the gist of the present specification, such detailed description will be omitted. Furthermore, identical reference numerals are used for identical components in the drawings, and redundant descriptions of identical components will be omitted.

[0062]

[0063] Interchange between terms in the examples

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

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

[0066] - 'two or more', 'a plurality of', 'multiple', 'multiple'. (In embodiments, 'one or more' or 'at least one' may be further limited to 'two or more', 'plural', 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 decryption', 'decoding', 'decoding'

[0075] - 'coding', 'encoding and / or decoding'

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

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

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

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

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

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

[0082] - 'pixel', 'pixels', 'samples', 'pels'

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

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

[0085] - 'quad', 'quarternary'

[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., neighboring block, adjacent block, and surrounding block)

[0090] - 'collocated', 'collected'

[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-screen mode', 'Inter-screen prediction mode'

[0098] - 'Motion vector', 'Predicted motion vector', 'Advanced Motion Vector Prediction (AMVP)'

[0099] - 'list', 'candidate list'

[0100] - 'Spatial candidate', 'Spatial merge candidate'

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

[0102] - 'Prediction motion vector candidate', 'motion vector predictor'

[0103] - 'Prediction method', 'Prediction mode'

[0104] - 'Intra', 'Intra'

[0105] - 'Intra prediction', 'Intra prediction'

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

[0107] - 'Dequantization', 'scaling'

[0108] - 'Quantization matrix', 'Scaling list'

[0109] - 'Quantization matrix coefficients', 'matrix coefficients'

[0110] - 'Transform coefficient level', 'quantized level', 'quantized coefficient', 'quantized transform coefficient', 'quantized transform coefficient level'

[0111] - 'Dequantized coefficient', 'dequantized transform coefficient'

[0112] - 'Scanning type', 'Scanning direction'

[0113] - 'Directional mode', 'Angle mode', 'Angular mode', 'Intra prediction mode'

[0114] - '(mode) number of intra prediction mode', '(mode) index of intra prediction mode', '(mode) value of intra prediction mode', '(mode) angle of intra prediction mode', '(mode) direction of intra prediction mode', '(mode) number of intra prediction direction', '(mode) index of intra prediction direction', '(mode) value of intra prediction direction', '(mode) angle of intra prediction direction'

[0115] - 'Merge Mode', 'Movement Merge Mode'

[0116] - 'Geometric Partitioning Mode (GPM)', 'Triangle Partitioning Mode'

[0117] In addition to the terms exemplified above, terms having the same meaning according to common knowledge in the technical field may be used interchangeably in the embodiments.

[0118]

[0119] The range of information and values ​​of information described in the examples

[0120] In embodiments, information may include constants, flags, indices, variables, coding parameters, elements, syntax elements, motion information, attributes, entities, objects, and data. That is, the term 'information' may be replaced with 'data', 'flag', 'index', 'variable', 'element', 'syntax element', 'motion information', 'attribute', or 'object'.

[0121] Information can have one of multiple values. The 'nth value' can mean the nth value among the multiple values.

[0122] For example, the first value could represent '0' or (logical) false. The second value could represent '1' or (logical) true. Alternatively, the first value could represent '1' or (logical) true. The second value could represent '0' or (logical) false.

[0123] A flag may be information having a value of either '0' or '1'. In embodiments, the flag values ​​'0' and '1' may be replaced with '1' and '0', respectively. For example, information indicating whether a specific process is performed or whether a specific process is applied may be considered a flag.

[0124] When a variable such as i or j is used to represent a row, column, or index, the variable can be an integer greater than or equal to 0 and less than or equal to n - 1. Alternatively, the variable can be an integer greater than or equal to 1 and less than or equal to n. Here, n can be the number of rows, the number of columns, or the number of entities pointed to by the index.

[0125]

[0126] Coding related concepts

[0127] Below, concepts related to coding are described. The descriptions disclosed below can be applied to embodiments.

[0128] Predefined value: A predefined value may refer to a value commonly used in an encoding device and a decoding device. For example, a predefined value may be interpreted as a fixed value. Alternatively, the predefined value may be a value shared by an encoding device and a decoding device through signaling. Alternatively, the predefined value may be a value derived through the same procedure in an encoding device and a decoding device so that the encoding device and the decoding device have a common value. Alternatively, the predefined value may be a common value in an encoding device and a decoding device. The above description of a predefined value may also be applied to predefined information. In the above descriptions, 'value' may be replaced with 'information'.

[0129] Availability: The availability of certain modes for a specific target may mean that a selected mode among the specific modes is used for the specific target. Other modes within the specific mode category may be unavailable. Unavailable modes may not be used for the specific target. The description of a specific mode above may also apply to other specific information. In the descriptions above, "mode" may be replaced with "information."

[0130] Adjacency: The 'direction' of the 'second entity' with respect to the 'first entity' may refer to the 'second entity' that is adjacent to the 'direction' corner / face of the first entity. For example, the 'top left block' with respect to the 'target block' may be a block adjacent to the top left of the target block. Here, the 'first entity' may be a target unit, a target block, or a target sample. The 'direction' may be one of left-above, above, right-above, left, right, left-below, below, and right-below. The 'second entity' may be a unit, a block, or a sample. For the directions of left-top, right-top, left-bottom, and right-bottom, the corners of the first entity and the corners of the second entity may be diagonally adjacent. For the directions of top, left, right and bottom, one side of the first object and one side of the second object can be in contact with each other.

[0131] - For example, the block adjacent to the upper left of the target block may be the block adjacent to the upper left of the block adjacent to the target block. The block adjacent to the upper right of the target block may be the block adjacent to the right of the block adjacent to the upper right of the target block. The block adjacent to the lower left of the target block may be the block adjacent to the lower left of the block adjacent to the target block.

[0132] Coding: Coding can mean encoding and / or decoding of images.

[0133] 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.

[0134] Video: A video can refer to a single picture that constitutes a video, or it can refer to the video itself. For example, "encoding and / or decoding a video" can mean "encoding and / or decoding a video," or it can mean "encoding and / or decoding one of the pictures that constitute the video."

[0135] - A picture can mean the entire picture, or it can mean a part of a picture, such as a block.

[0136] Target Image: The target image may be an encoding target image, which is the target of encoding, and / or a decoding target image, which is the target of decoding. Furthermore, the target image may be an input image processed by an encoding device, or a restored image processed by a decoding device. The target image may be an image including a target block.

[0137] Subpicture: A picture can be divided into one or more subpictures.

[0138] - A subpicture may be a square or rectangular area within a picture. A subpicture may contain one or more CTUs.

[0139] - 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.

[0140] - A subpicture may include one or more slices that collectively cover a rectangular area within the picture. Accordingly, the boundary of each subpicture may always be the boundary of a slice. Additionally, each vertical subpicture boundary may always be a vertical tile boundary.

[0141] Slice: A slice may contain 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.

[0142] 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.

[0143] CTU: An image can be divided into multiple coding tree units (CTUs).

[0144] - A CTU may include one Y coding tree block (CTB) and at least one of a Cb CTB and a Cr CTB related to the Y CTB, and may include information about each CTB. The information may include syntax elements.

[0145] - Each CTU can be partitioned using one or more partitioning methods to form sub-units such as coding units (CUs), prediction units (PUs), and transform units (TUs). The one or more partitioning methods can include quad tree (QT) partitioning, binary tree (BT) partitioning, and ternary tree (TT) partitioning. Additionally, each CTU can be partitioned using multi-type tree (MTT) partitioning that uses a combination of multiple partitioning methods.

[0146] CTB: CTB can refer to one of Y CTB, Cb CTB, and Cr CTB.

[0147] Unit: A unit can be determined for specific processing in coding. A unit can contain information about a specific region within an image. For specific coding processing, an image can be recursively divided into multiple parts. A unit can represent the region to which a specific processing is applied and information about the region.

[0148] - The type of a unit may indicate a specific processing to be applied to the unit. Depending on the type of the unit, a specific processing may be applied to the unit. A 'specific' unit may be a unit for processing designated as 'specific' in coding. For example, the unit may be at least one of an original unit, a CTU, a coding unit, a prediction unit, a residual unit, a reconstructed residual unit, a transformation unit, and a reconstructed unit.

[0149] - A unit may include samples having a two-dimensional shape or arrangement. In this respect, a 'unit' may also 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 reconstructed residual block, a transform block (TB), and a reconstructed block. For example, a division of a unit may mean a division of a block corresponding to the unit.

[0150] - A unit can contain syntax elements. In other words, a block and its syntax elements can be combined to form a unit.

[0151] - A block is an MxN array of samples. Here, M and N can represent positive integer values, and a block can commonly represent a two-dimensional sample array. The current block can represent an encoding target block that is the target of encoding during encoding, and a decoding target block that is the target of decoding during decoding. In addition, the current block can be at least one of a coding block, a prediction block, a residual block, a transform block, and a restoration block. The block can have various sizes and shapes. For example, the shape of the block can be one or more of a tetragon, a rectangular block, a square block, a rectangle whose width is different from its height (that is, an oblong block), a trapezoid, a triangle, a right-angled triangle, and a pentagon. Here, the width and height of the rectangle can be different from each other. In addition, the shape of the block can include other geometric shapes that can be expressed in two dimensions. For example, the shape of a block may be a quadrilateral or a pentagon, which is defined by excluding the area of ​​a right triangle from the area of ​​a rectangle. Here, the right vertex of the right triangle may be one of the vertices of the rectangle. Furthermore, the shape of a block may be a combination of two or more of the aforementioned shapes. Furthermore, the shape of a block may be the remainder of one of the aforementioned shapes after excluding another shape.

[0152] - In embodiments, a rectangle may be limited to a non-square rectangle. When the shape of a particular object is described as a rectangle in an embodiment, such description may additionally imply that the width and height of the particular object are different from each other.

[0153] - In embodiments, a 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 whose vertical length is greater than its horizontal length. A horizontally oriented block may mean a block whose horizontal length is greater than its vertical length.

[0154] - A 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 about each block. The information may include syntax elements.

[0155] - Unit information may include unit type, unit size, unit depth, unit encoding order, and unit decoding order.

[0156] Target Unit: A target unit may be a block, an encoding target unit, which is a target of encoding, and / or a decoding target unit, which is a target of decoding. A target unit may be a specific area within a target picture to which one or more specific coding processes are applied. A unit of a specific type may be generated by applying a specific process to a target unit. Alternatively, a target unit may represent a unit having a specific type for a specific coding process.

[0157] Depth: A block can be hierarchically divided into multiple sub-blocks, each with its own depth, according to a tree structure. The multiple sub-blocks created by block division can be called partitions.

[0158] - The depth of a block can indicate the level of the node corresponding to the block when the blocks that make up the image are expressed in a tree structure. Alternatively, the depth of a block can indicate the number of partitions applied until the block is determined. The depth of a block can increase by 1 as the block is further partitioned.

[0159] - In a tree structure, the root node can be considered to have the smallest level, and the leaf node can be considered to have the largest level. The root node can be the topmost node in the tree structure and corresponds to the first undivided block. The level of the root node can be 0 or 1. When the level of the root node is 0, a node with a level of 1 can represent a block determined by dividing the first block once. A node with a level of n can represent a block determined by dividing the first block n times. A leaf node can be the lowest node in the tree structure. A leaf node can be a node that cannot be further divided. The depth of a leaf node can be a predefined maximum depth. For example, the maximum depth can be a positive integer such as 3. The root node can mean a CTU. A leaf node can mean at least one of a CU, a PU, and a TU.

[0160] - 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.

[0161] Sample: A sample can be a base unit that constitutes a block. A sample can be composed of one or more bits. The bit depth can be the number of bits that constitute a sample. A sample can be numbered from 0 to 2 depending on the bit depth. Bd It can be expressed as values ​​up to -1.

[0162] PU: PU may denote a basic unit for prediction-related processing. For example, prediction-related processing may include inter-prediction, intra-prediction, intra-block copy (IBC) prediction, intra-compensation, and motion compensation.

[0163] - A PU can be divided into multiple sub-PUs, each of which has a smaller size than the PU itself. These multiple sub-PUs can also serve as the basis for prediction-related processing. In other words, a prediction unit partition generated by splitting a prediction unit can also be a prediction unit.

[0164] TU: TU may be a basic unit for processing related to a residual block. The processing related to the 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.

[0165] - A TU can be divided into multiple sub-transformation units with sizes smaller than the TU size. The multiple sub-TUs can also serve as basic units for processing related to residual blocks. In other words, a transform unit partition generated by splitting a transform unit can also be a transform unit.

[0166] - 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.

[0167] Parameter set: A parameter set may correspond to header information among the structures within a bitstream.

[0168] - 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).

[0169] - Information signaled through a parameter set can be applied to pictures referencing the parameter set. For example, information within a VPS can be applied to pictures referencing the VPS. Information within an SPS can be applied to pictures referencing the SPS. Information within a PPS can be applied to pictures referencing the PPS. A parameter set can refer to a higher-order parameter set. For example, a PPS can refer to an SPS. An SPS can refer to a VPS.

[0170] - Additionally, the parameter set may include tile group information, slice header information, and tile header information. A tile group may mean a group or slice including multiple tiles.

[0171] MPM (Most Probable Mode): MPM can indicate the intra prediction mode that is likely to be used for intra prediction for the target block.

[0172] - One or more different MPMs can be determined based on coding parameters related to the target block and properties of objects related to the target block.

[0173] - One or more MPMs may be determined based on the intra prediction mode of a reference block. There may be multiple reference blocks. Depending on which intra prediction modes are used for one or more reference blocks, one or more different MPMs may be determined. The reference blocks may include spatial neighboring blocks.

[0174] MPM List: An MPM list may contain one or more MPMs. The number of MPMs in an MPM list may be predefined.

[0175] MPM Index: The MPM index can indicate an MPM among one or more MPMs in the MPM list to be used for intra prediction for the target block.

[0176] MPM Usage Directive: The MPM usage directive can indicate whether the MPM list is used for prediction on the target block.

[0177] 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. In addition, the prediction mode may include at least one of an intra mode, an inter mode, and an intra block copy mode.

[0178] Reference image list: The reference image list may be a list containing one or more reference images used for prediction for the target block.

[0179] - There may be multiple reference image lists. Multiple reference image lists may include List 0 (L0), List 1 (L1), etc.

[0180] - One or more reference image lists may be used for inter prediction for a target block. Parts such as 'L0' and 'L1' in the names of information related to inter prediction may refer to reference image lists related to the information.

[0181] Reference picture: A reference picture may be an image referenced for prediction of a target block. Alternatively, the reference picture may be an image containing a reference block. The reference picture may include a previous image of the target image, a target image, and a subsequent image of the target image.

[0182] 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.

[0183] Reference Block: A reference block may be a block referenced for encoding / decoding a target block, such as prediction or filtering. For example, a reference block may include reference samples used to derive prediction samples, and may also refer to a block that provides information used for decoding the target block.

[0184] Reference sample: A reference sample may be a sample that is referenced for encoding / decoding of a target block, such as prediction and filtering.

[0185] Inter prediction indicator: The inter prediction indicator can indicate the direction of inter prediction for the target block. The inter prediction can be one of uni-directional prediction and bi-directional prediction. Alternatively, the inter prediction indicator can indicate the number of reference pictures used when generating a prediction block of the target block. Alternatively, the inter prediction indicator can indicate the number of prediction blocks used for inter prediction for the target block. The reference direction can mean the inter prediction indicator. For example, the inter prediction indicator can indicate one of uni-directional and bi-directional. Alternatively, the inter prediction indicator can have a first value of '0' for an inter mode that uses only reference pictures in the L0 reference picture list, a second value of '1' for an inter mode that uses only reference pictures in the L1 reference picture list, and a third value of '2' for an inter mode that uses at least two of the reference pictures in the L0 reference picture list and the reference pictures in the L1 reference picture list.

[0186] Prediction List Utilization Flag: The prediction list utilization flag for a specific reference image list may indicate whether at least one reference image within the specific reference image list is used to generate a prediction block of the target block. For example, a value of the prediction list utilization flag for a specific reference image list of '0' may indicate that a prediction block is not generated using a reference image within the specific reference image list. A value of the prediction list utilization flag for a specific reference image list of '1' may indicate that a prediction block is generated using a reference image within the specific reference image list.

[0187] - An inter prediction indicator can be derived using a prediction list utilization flag. Conversely, a prediction list utilization flag can be derived using an inter prediction indicator. For example, an inter prediction indicator can be derived using prediction list utilization flags for a plurality of reference image lists. If an inter prediction indicator indicates that specific reference lists among a plurality of reference image lists are used, the prediction list utilization flags of the specific reference lists indicated by the inter prediction indicator among the prediction list utilization flags of the plurality of reference image lists can be set to '1', and the prediction list utilization flags of the remaining reference image lists not indicated by the inter prediction indicator can be set to '0'.

[0188] 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 the reference image list L0 and the reference image list L1.

[0189] - The reference direction may not indicate that the directions of the reference images in the reference image list are limited to the forward direction or the backward direction, but only indicates the reference image list used for prediction of the target block. That is, each of the reference image list L0 and the reference image list L1 may include forward images and backward images. Here, the forward direction may indicate the direction from the target image to the previous image of the target image. Forward inter prediction may be inter prediction that uses the previous image of the target image as a reference image. Backward direction may indicate the direction from the target image to the subsequent image of the target image. Backward inter prediction may be inter prediction that uses the subsequent image of the target image as a reference image.

[0190] - A unidirectional reference direction may mean that one 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 that only the reference image list L0 is used, that only the reference image list L1 is used, or that two reference image lists are used. Additionally, the reference direction may be indicated by an inter prediction indicator.

[0191] Picture Order Count (POC): The POC of a picture can indicate the display order or output order of the picture.

[0192] Motion information: Motion information may be information used to specify a reference block. Motion information may include information used for inter prediction, such as a motion vector (MV), a reference picture index, a reference picture, an inter prediction indicator, and a prediction list utilization flag. Additionally, motion information may include information used in a specific inter prediction mode, such as an MV candidate, an MV candidate index, a merge candidate, and a merge index.

[0193] - For inter prediction of a target block, multiple motion information for multiple reference image lists can be used, respectively. Motion information for a specific reference image list can be used for prediction using the specific reference image list. Multiple (intermediate) prediction blocks can be derived from the multiple motion information. A (final) prediction block for the target block can be generated using statistical values ​​for the multiple (intermediate) prediction blocks.

[0194] MV: MV can be a two-dimensional vector used in inter prediction. It can represent the offset between a target block and a reference block. Alternatively, it can represent the difference between the locations of a target block and a reference block.

[0195] - For example, MV is (mv x , mv y ) can be expressed in the form of mv x can represent the horizontal component, and mv y can represent vertical components.

[0196] -The zero vector can be (0, 0) MV.

[0197] 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 the target image. In other words, a BV can represent the displacement between the target block and the reference block within the target image.

[0198] - For example, BV is similar to MV (bv x , bv y ) can be expressed in the form of bv x can represent the horizontal component, bv y can represent vertical components.

[0199] -The zero vector can be (0, 0) BV.

[0200] Motion Information Candidate: In a specific prediction, the motion information of the target block can be selected from among motion information candidates determined by a specific method. The motion information candidate may refer to the motion information of a reference block, or it may refer to the reference block itself containing the motion information. Here, the reference block may be a block determined by a specific method for selecting a motion information candidate.

[0201] Candidate List: A candidate list may be a list containing one or more candidates. For example, the candidate list may include a motion information candidate list, a merge candidate list, an MV candidate list, an MPM list, etc. The candidate list may be generated in the same manner by the encoding device and the decoding device. In other words, the candidate list used by the encoding device and the candidate list used by the decoding device may be the same, and the same candidate list may be shared by the encoding device and the decoding device. The encoding device may select a candidate to be used for processing the target block from among the candidates in the candidate list. An indicator indicating the selected candidate may be signaled from the encoding device to the decoding device. The decoding device may use the indicator to specify a candidate to be used for processing the target block from among the candidates in the candidate list. Alternatively, the encoding device and the decoding device may specify a candidate to be used for processing the target block from among the candidates in the candidate list according to the same rule.

[0202] Motion information candidate list: The motion information candidate list may mean a list constructed using one or more motion information candidates.

[0203] Motion information candidate index: The motion information candidate index may be an identifier or indicator that indicates a motion information candidate used for prediction of a target block among the motion information candidates in the motion information candidate list.

[0204] - In a specific inter prediction mode, motion information of other reconstructed blocks may be used to derive motion information of the target block. The other blocks may include neighboring blocks. In this specific inter prediction mode, the motion information for the target block itself is not individually signaled, but other information used to derive motion information of the target block based on the motion information of other reconstructed blocks may be signaled. In this case, the other information may include information indicating which of the other reconstructed blocks' motion information is used to derive motion information of the target block, such as a motion information candidate index.

[0205] - 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.

[0206] - In embodiments, MV may be part of motion information. In embodiments, information about motion information, such as motion information candidates, motion information candidate lists, and motion information candidate indices, may be replaced with information about MVs, such as MV candidates, MV candidate lists, and MV candidate indices, and the description of motion information may also be applied to MVs.

[0207] Merge: Merge can refer to the merging of motion information across multiple blocks, or it can refer to applying motion information from another block to the target block. In other words, merge mode can refer to a mode in which the motion information of the target block is derived from the motion information of neighboring blocks.

[0208] Merge Candidate: A merge candidate may refer to a specific (restored) block used for merging the target block, or may refer to motion information for the specific block. Alternatively, the merge candidate may include motion information for the specific block.

[0209] - 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.

[0210] Merge Candidate List: A merge candidate list may be a list constructed using one or more merge candidates.

[0211] Merge Index: A merge index may be an indicator that points to a merge candidate among the merge candidates in the merge candidate list, which is used for prediction of the target block. The motion information of the merge candidate indicated by the merge index among the merge candidates in the merge candidate list may be used as motion information of the target block.

[0212] Neighboring block: A neighboring block can refer to a block adjacent to the target block. Neighboring blocks can include spatial and temporal neighboring blocks. A neighboring block can also refer to a reconstructed neighboring block within a reference image.

[0213] Spatial neighboring blocks: Spatial neighboring blocks can be blocks that are spatially adjacent to the target block.

[0214] - The target block and spatial neighboring blocks can be included within the target image.

[0215] - A spatial neighboring block may include a block whose boundary is at least partially adjacent to a boundary of the target block. Alternatively, a spatial neighboring block may include a block whose distance from the target block is less than or equal to a specific value.

[0216] - A spatial neighboring block may include a block diagonally adjacent to a vertex of the target block.

[0217] - Spatial neighboring blocks may include an upper left block adjacent to the upper left of the target block, an upper block adjacent to the upper right of the target block, an upper right block entered at the upper 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 lower left block adjacent to the lower left of the target block, a lower block adjacent to the lower bottom of the target block, and a lower right block adjacent to the lower right of the target block.

[0218] Temporal neighboring blocks: Temporal neighboring blocks can be blocks that are temporally adjacent to the target block.

[0219] - A temporal neighboring block may include a collocated block (COL block). A collocated block may be a block within a reconstructed image within a reference image buffer. A collocated picture (col picture) may refer to an image that includes a collocated block. A collocated picture may be an image included in a reference image list.

[0220] - Call blocks can be determined based on the location of the target block within the target image. Two blocks being "temporally adjacent" can mean that the locations of the two blocks satisfy certain conditions.

[0221] - The position of a call block within a call image may be the same as the position of a target block within a target image. Alternatively, the position of a call block within a call image may correspond to the position of a target block within a target image. Here, the correspondence of the positions of blocks may mean that the areas of the blocks are identical, that an area of ​​one block is included in an area of ​​another block, or that one block occupies a specific position of another block.

[0222] - For example, the location of a call block within a call image may be identical to the location of a target block within the target image. Alternatively, a call block may be a block containing a call sample 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.

[0223] - A temporal neighboring block may be a block that is temporally adjacent to a spatial neighboring block of the target block.

[0224] Search range: The search range can refer to a two-dimensional region where MVs are searched during inter prediction. For example, when the optimal MV must be derived for processing a target block, the optimal MV can be selected from among the MVs pointing within the search range.

[0225] 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 dequantization on a quantized level.

[0226] Quantized level: A quantized level can be an integer quantity used as input to dequantization.

[0227] Quantization: Quantization can be the process of generating quantized levels for transform coefficients. Quantized levels can be generated by applying quantization to transform coefficients. The transform can also be considered part of quantization.

[0228] Dequantization: Dequantization can be the process of multiplying a quantized level by a factor. By applying dequantization to a quantized level, (restored) transform coefficients can be generated.

[0229] Quantization Parameter (QP): QP can refer to an argument used to generate quantized levels for transform coefficients in quantization. QP can also refer to an argument used to generate (restored) transform coefficients for quantized levels in dequantization. Alternatively, QP can be a value mapped to the quantization step size.

[0230] 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 the delta QP.

[0231] Quantization matrix: A quantization matrix can be a matrix used in quantization or inverse quantization to improve the subjective or objective quality of an image.

[0232] Quantization matrix coefficients: Quantization matrix coefficients can be each element within a quantization matrix.

[0233] Scan: A scan can refer to the arrangement of values ​​within a block or matrix. The values ​​can be coefficients. For example, a scan can refer to arranging values ​​arranged in a two-dimensional form into a one-dimensional form, or it can refer to rearranging values ​​arranged in a one-dimensional form into a two-dimensional form. An inverse scan can be the opposite arrangement (or rearrangement) of the arrangement performed in a scan.

[0234] Non-zero transform coefficient: A non-zero transform coefficient can mean a transform coefficient with a non-zero value or a quantized level with a non-zero value.

[0235] Bitstream: A bitstream may refer to a sequence of bits containing encoded information generated by encoding an image. A bitstream may contain information according to specific syntax elements. For example, information may contain syntax elements. An encoding device may generate a bitstream containing information according to specific syntax elements. A decoding device may obtain information from the bitstream according to specific syntax elements.

[0236] Signaling: Signaling information may indicate that information is transmitted from an encoding device to a decoding device via a bitstream. For example, the information may include a syntax element. Alternatively, signaling may mean that the encoding device includes information in a bitstream. Information signaled by the encoding device may be used by the decoding device. In signaling, the bitstream may be transmitted via a network and may be included in a recording medium. In embodiments, description of information being signaled may include: 1) for signaling information, the encoding device determines and generates information; 2) the encoding device encodes the information to generate encoded information; 3) (encoded) information is transmitted from the encoding device to the decoding device via a bitstream; 4) the decoding device decodes the encoded information to obtain information; and 5) for signaling information, the decoding device determines and generates information via signaling.

[0237] - An encoding device can perform encoding on information to generate encoded information. The encoded information can be signaled via a bitstream. A decoding device can obtain information by decoding the encoded information.

[0238] - When information is signaled for a specific target, it can mean that the information is used for each specific target, and the processing indicated by the information is applied to each specific target. For example, when information is signaled at a specific unit level, it can mean that the information is used / processed for each specific unit.

[0239] - The information being signaled may include one or more sub-information. Signaling a specific piece of information may mean that each piece of information within one or more sub-information pieces contained within the specific information is signaled.

[0240] Selective Signaling: Signaling of information may be performed selectively. Selective signaling of information may mean that the encoding device selectively includes information in the bitstream (under certain conditions). Selective signaling of information may mean that the decoding device selectively obtains information from the bitstream (under certain conditions).

[0241] Omission of signaling: Signaling for information may be omitted. Omission of signaling for information may mean that the encoding device (under certain conditions) does not include the information in the bitstream. Omission of signaling for information may mean that the decoding device (under certain conditions) does not obtain the information from the bitstream. The decoding device may derive the information for which signaling is omitted using other information of the embodiments.

[0242] Symbol: may mean at least one piece of information of a target unit, such as a syntax element of a target unit or target block, a coding parameter, a quantized level, and a transform coefficient. In addition, a symbol may mean a target of entropy encoding or a result of entropy decoding.

[0243] Entropy encoding: Entropy encoding can allocate fewer bits to symbols with a high probability of occurrence, and more bits to symbols with a low probability of occurrence. This allocation reduces the size of the bitstream representing the symbols as they are represented.

[0244] - Entropy coding can use 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 a variable length table. For example, in CABAC, a binarization method for symbols and a probability model of symbols / bins can be derived for entropy coding, and arithmetic coding using context can be performed.

[0245] Entropy decoding: Entropy decoding can reverse the processes performed in entropy encoding. Symbols can be generated by entropy decoding a bitstream.

[0246] Parsing: Parsing can mean determining the values ​​of syntactic elements by performing entropy decoding on the encoded information in the bitstream. Alternatively, parsing can mean entropy decoding itself.

[0247] Statistical Value: The values ​​of information related to specific entities described in the embodiments may be used as inputs to specific operations. The statistical value may be a value derived by a specific operation on the values ​​related to these specific entities. For example, the statistical value for specific information may be one or more of an average value, a weighted average value, a weighted sum value, a minimum value, a maximum value, a mode, a median value, an interpolated value, a sum of products, and a product of sums of values ​​of the specific information. Additionally, information of the embodiments having specific values ​​determined by operations, such as constants, variables, and coding parameters, may have specific statistical values ​​according to the embodiments.

[0248]

[0249] Coding parameters

[0250] In embodiments, coding parameters may be information required for coding. The coding parameters may include information signaled from an encoding device to a decoding device, information calculated / derived during the coding process described in the embodiments, and information used for the coding process described in the embodiments.

[0251] In embodiments, the coding parameters include a size of a CTU, a size of a unit, a form of a unit, a shape of a unit, a depth of a unit, a minimum unit size, a maximum unit size, a maximum unit depth, a minimum unit depth, a partition information of a unit, QT partition information, BT partition information, a partition direction of a BT partition, a partition shape of a BT partition, TT partition information, a partition direction of a TT partition, a partition shape of a TT partition, MTT partition information, a combination of MTT partitions, a partition direction of an MTT partition, a partition shape of an MTT partition, a prediction mode, an intra prediction mode, a luma intra prediction mode, a chroma intra prediction mode, an intra partition information, an inter partition information, a coding block partition information, a prediction block partition information, a transform block partition information, a reference sample line index, a reference sample filtering method, a reference sample filter tap, a reference sample filter coefficient, a prediction block filter method, a prediction block filter tap, a prediction block filter coefficient, a prediction block boundary filtering method, a prediction block boundary filter tap, a prediction block boundary filter coefficient, an inter prediction mode, motion information, MV, a motion vector difference (MV). Difference (MVD), MVD resolution, MV size, MV representation accuracy, reference picture list, reference picture, reference picture 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 compensation 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, filter tap of interpolation filter, filter coefficient of interpolation filter, transformation type, transformation size, transformation selection information, primary transformation 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, coefficients of the deblocking filter, filter taps of the deblocking filter, strength of the deblocking filter, shape / shape of the deblocking filter, adaptive sample offset usage information, adaptive sample offset value, adaptive sample offset category, adaptive sample offset type, adaptive loop filter usage information, coefficients of the adaptive loop filter, filter taps of the adaptive loop filter, shape / shape of the adaptive loop filter, 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 unit coding flag, last significant coefficient position, flag indicating whether the coefficient value is greater than 1, whether the coefficient value is greater than 2 A flag indicating whether the coefficient value is greater than 3, a flag indicating whether the coefficient value is greater than 3, remaining coefficient value information, sign information, context bin, bypass bin, reconstructed sample, reconstructed luma sample, reconstructed chroma sample, residual sample, residual luma sample, residual chroma sample, transform coefficient, luma transform coefficient, chroma transform coefficient, transform coefficient level, luma transform coefficient level, chroma transform coefficient level, transform coefficient level scanning method, quantized level, luma quantized level, chroma quantized level, size of MV search region on the side of the decoding device, shape of MV search region on the side of the decoding device, number of MV search on the side of the decoding device, picture type, slice identification information, slice type, slice partitioning information, tile group identification information, tile group type, tile group partitioning information, tile identification information, tile type, tile partitioning information, bit depth,It may include one or more of input sample bit depth, reconstructed sample bit depth, residual sample bit depth, transform coefficient bit depth, quantized level bit depth, mapping availability information, information about luma signal, information about chroma signal, color space of target block, color space of residual block, and temporal layer information.

[0252] 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 for 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.

[0253] In embodiments, "X usage information" may be "information indicating whether X is used / applied / performed." 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. The mode information may indicate a mode used for a target block among the modes described in the embodiments. In embodiments, the specific mode usage information may be replaced with mode information, and the description of the specific mode usage information may also be applied to the mode information. "X usage information" and "X indicator" may be used interchangeably.

[0254] In embodiments, coding parameters and syntax elements may correspond to each other. For example, syntax elements of an embodiment may be used as coding parameters, and coding parameters may be signaled as syntax elements.

[0255] In embodiments, “X presence information” may be considered as “information indicating whether X exists” or “information indicating whether information indicating X exists in the bitstream.”

[0256] In embodiments, the “X selection information” may be information indicating one of the candidates or methods for X. The “X selection information” may be considered an “X index.”

[0257] In embodiments, the splitting form of a particular 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 particular tree may represent one of horizontal and vertical directions.

[0258] In embodiments, when a coding parameter has one of multiple values, “coding parameter” may be replaced with “whether the coding parameter has a specific value among the multiple values ​​available to the coding parameter.”

[0259] In embodiments, when a coding parameter points to one of a plurality of objects, “coding parameter” may be replaced with “whether the coding parameter points to a specific object among the plurality of objects.”

[0260]

[0261] System for video coding

[0262] Figure 1 illustrates a system for video coding according to one embodiment.

[0263] The system (100) may include at least one of an encoding device (110) and a decoding device (150).

[0264] Each of the encoding device (110) and the decoding device (150) may be a computer or an electronic apparatus.

[0265]

[0266] Structure of the encoding device

[0267] The encoding device (110) may include a processor (120), storage (140), and a communicator (149).

[0268] The processor (120), storage (140), and communication device (149) can be connected via a bus.

[0269] The processor (120) may be a semiconductor device that executes instructions or computer-executable codes, such as a central processing unit (CPU). The processor (120) may be at least one hardware processor.

[0270] The processor (120) can perform generation and processing of information input to the encoding device (110), output from the encoding device (110), or used within the encoding device (110) in the embodiments, and can perform comparisons and judgments related to such information.

[0271] 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).

[0272] At least some of the aforementioned components may be program modules. The program modules may be included in the encoding device (110) in the form of an operating system, applications, and other program modules. The program modules may be instructions or computer-executable codes stored in the storage (140) and executed by the processor (120).

[0273] 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.

[0274] The storage (140) can store instructions and computer-executable codes used for the operation of the encoding device (110), and can store information and bitstreams described in the embodiments. The storage (140) can include a reference picture buffer (141).

[0275] 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).

[0276] Among the names of components of the encoding device (110), “-er” or “-or” may be replaced with “-unit”. The storage (140) may also be named a storage unit.

[0277]

[0278] Operation of the encoding device

[0279] The encoding device (110) can sequentially encode one or more images of a video.

[0280] The storage (140) can store the original image. The original image can be used as a target image in the encoding device (110).

[0281] The processor (120) can generate a bitstream including encoded information by performing encoding on the target image, and can store the generated bitstream in the storage (140). The generated bitstream can be stored in a computer-readable recording medium, and can be transmitted to the communication device (189) of the decoding device (150) via a wired and / or wireless transmission medium by the communication device (149).

[0282] The segmenter (122) can determine a target block by performing segmentation on the target image.

[0283] 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 prediction according to the prediction mode.

[0284] The prediction mode of the target block may be one of the available prediction modes. For example, the available prediction modes may include intra prediction, inter prediction, and IBC prediction.

[0285] 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.

[0286] 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.

[0287] For example, when the prediction mode is IBC, the predictor (123) can perform IBC prediction on the target block to generate a prediction block of the target block.

[0288] 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 predicted block. The original block may be the area pointed to by the target block in the original image. 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 predicted block.

[0289] The transformer (125) can perform a transformation on the residual block to generate transformation coefficients.

[0290] The converter (125) can perform the conversion using one of a plurality of conversion methods.

[0291] For example, the multiple transform methods may include a Discrete Cosine Transform (DCT), a Discrete Sine Transform (DST), a Karhunen-Loeve Transform (KLT), and transforms based on each transform.

[0292] Transform skip mode may be a mode for generating a reconstructed block using a reconstructed residual block and a prediction block for which transformation and inverse transformation have not been performed. When transform skip mode is applied to a target block, transformation and inverse transformation for the target block may be omitted, and only quantization and inverse quantization for the target block may be performed.

[0293] A quantizer (126) can generate quantized levels by applying quantization using quantization parameters to transform coefficients. In embodiments, the quantized levels may also be referred to as transform coefficients.

[0294] An entropy encoder (139) can generate encoded information by performing entropy encoding based on a probability distribution on information for decoding an image. The bitstream can include encoded information.

[0295] Information for decoding an image may include quantized levels and syntax elements produced by a quantizer (126).

[0296] The probability distribution can be determined based on the quantized levels and coding parameters.

[0297] The entropy encoder (139) can use scanning to change the quantized levels in the form of two-dimensional blocks into the form of one-dimensional vectors in order to perform encoding on the quantized levels. In scanning, which scan among the upper right diagonal scan, vertical scan, and horizontal scan will be used can be determined based on coding parameters such as the block size and the block intra prediction mode.

[0298] 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 to be processed later. Accordingly, the processor (120) can perform restoration on the encoded target block, and store a restored image including the restored target block generated by the restoration as a reference image in the reference picture buffer (141). Inverse quantization and inverse transformation can be performed on the encoded target block for restoration.

[0299] The dequantizer (127) can generate dequantized transform coefficients by performing dequantization on the quantized level.

[0300] The inverse transformer (128) can generate inverse quantized and inversely transformed coefficients by performing inverse transformation on the inverse quantized transform coefficients. In embodiments, the inverse quantized and / or inversely transformed coefficients may refer to coefficients to which at least one of inverse quantization and inverse transformation has been applied. The inverse quantized and inversely transformed coefficients may be a restored residual block.

[0301] The adder (129) can generate a restored block by combining a predicted block and a restored residual block.

[0302] 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 filter 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.

[0303] The reference picture buffer (141) can store a restored block / image provided from the filter (130). The restored image may be an image including a restored block. Alternatively, the restored image may be an image composed of restored blocks.

[0304] The reference picture buffer (141) can provide the stored restored image as a reference image to the predictor (123). In terms of storing the decoded (i.e., restored) picture, the reference picture buffer (141) may also be referred to as a decoded picture buffer (DPB).

[0305]

[0306] Structure of the decryption device

[0307] The decryption device (150) may include a processor (160), a storage (180), and a communication device (189).

[0308] The description of the processor (120), storage (140), and communication device (149) related to the encoding device (110) can also be applied to the processor (160), storage (180), and communication device (189) related to the decoding device (150). Duplicate descriptions are omitted.

[0309] The processor (160) may include a plurality of components. The plurality of components may include an entropy decoder (161), a divider (162), a predictor (163), an inverse quantizer (167), an inverse transformer (168), an adder (169), and a filter (170).

[0310] The storage (180) may include a reference picture buffer (181).

[0311] The communication device (189) can perform functions related to communication of information in the decryption device (150). For example, the communication device (189) can receive a bitstream from the encoding device (110).

[0312] Among the names of components of the decryption device (150), “-er” or “-or” may be replaced with “-unit”. The storage (180) may also be named a storage unit.

[0313]

[0314] Operation of the decryption device

[0315] The communication device (149) of the encoding device (110) can transmit the bitstream generated by the encoding device (100) to the decoding device (150). Alternatively, a computer-readable recording medium storing the bitstream can transmit the bitstream generated by the encoding device (100) to the decoding device (150).

[0316] 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 the storage (180).

[0317] The processor (160) can obtain a bitstream from a storage (180) or a computer-readable recording medium.

[0318] A bitstream may contain encoded information.

[0319] An entropy decoder (161) can generate information for decoding an image by performing entropy decoding based on a probability distribution on the encoded information of a bitstream.

[0320] Information for decoding an image may include quantized levels and syntax elements.

[0321] The entropy decoder (161) can use scanning to change the quantized levels in the form of a one-dimensional vector into the form of a two-dimensional block to perform decoding on the quantized levels. In scanning, which scan among the upper right diagonal scan, vertical scan, and horizontal scan will be used can be determined based on coding parameters such as the block size and the block intra prediction mode.

[0322] The entropy decoder (161) can provide syntax elements to other components of the processor (160), such as the segmenter (162).

[0323]

[0324] A common description of the relationship between the components of the encoding device and the components of the decoding device.

[0325] The decoding device (150) performs decoding using the bitstream generated by the encoding device (110). The encoding device (110) can perform encoding on 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). Therefore, the encoding device (110) and the decoding device (150) may need to generate restored blocks / images in the same manner. In this respect, the descriptions of the divider (122), predictor (123), inverse quantizer (127), inverse transformer (128), adder (129), filter (130), and reference picture buffer (141) of the encoding device (110) disclosed in the embodiments can also be applied to the divider (162), predictor (163), inverse quantizer (167), inverse transformer (168), adder (169), filter (170), and reference picture buffer (181) of the decoding device (150). Duplicate descriptions are omitted.

[0326] Additionally, each of the divider (122), predictor (123), inverse quantizer (127), inverse transformer (128), adder (129), and filter (130) of the encoding device (110) can generate syntax element information that specifies processing for the target. Each of the divider (162), predictor (163), inverse quantizer (167), inverse transformer (168), adder (169), and filter (170) of the decoding device (150) can perform processing for the target (same as that performed in the encoding device (110)) using the syntax element information.

[0327] 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, in the function related to 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). In the function related to transformation, the processing unit may represent a transformer (125) and an inverse transformer (128), and may represent an inverse transformer (168). In the function related to quantization, the processor may represent a quantizer (126) and an inverse quantizer (127), and may represent an inverse quantizer (167). In the function related to entropy encoding / decoding, the processing unit may represent an entropy encoder (139) and / or an entropy decoder (161). In the function related to filtering, the processing unit may represent a filter (130) and / or a filter (170). The storage may represent a storage (140) of an encoding device (110) and / or a storage (180) of a decoding device (150). The reference picture buffer may represent a reference picture buffer (141) of an encoding device (110) and / or a reference picture buffer (181) of a decoding device (150). The communication unit may represent a communication unit (149) of an encoding device (110) and / or a communication unit (189) of a decoding device (150).

[0328]

[0329] Division of the units that make up the image

[0330] Figure 2 shows a segmentation structure of an image according to one embodiment.

[0331] Figure 2 can schematically represent an example in which one unit is divided into multiple sub-units.

[0332] A CU can be used as a basic unit for encoding and decoding images. In addition, a CU can be a basic unit for prediction, transformation, quantization, inverse quantization, inverse transform entropy encoding, and entropy decoding.

[0333] A CU can be used as a unit to which a prediction mode is applied. In other words, during coding, it can be determined which prediction mode among 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 prediction (IBC).

[0334] 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, a single CTU can be used as a CU. The size of the CTU can be the maximum CU size.

[0335] Each CU can have depth information. The depth information can indicate the depth of the CU and the size of the CU. The depth of a CTU can be 0. The depth of a CU generated by splitting a CTU can be 1. When a parent CU is split into child CUs, the depth of the child CU can be 1 greater than the depth of the parent CU. The number of split CUs can be a positive integer greater than or equal to 2, including 2, 4, 8, and 16. At least one of the horizontal size and the vertical size of the child CU generated by splitting the parent CU can be smaller than at least one of the horizontal size and the vertical size of the parent CU, depending on the number of child CUs.

[0336] A partitioned CU can be recursively partitioned in the same manner up to a predefined maximum depth or a predefined minimum size. The depth of the smallest coding unit (SCU) can be the predefined maximum depth, and the size of the SCU can be the predefined minimum size. The size of the SCU can be the size of the minimum CU.

[0337] For example, the depth of a CU can range 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 a 64x64 block. 0 can be the minimum depth. An SCU with a depth of 3 can be an 8x8 block. 3 can be the maximum depth. A depth of 0 can represent a CTU that is a 64x64 block. A depth of 1 can represent a CU that is a 32x32 block. A depth of 2 can represent a CU that is a 16x16 block. A depth of 3 can represent an SCU that is an 8x8 block.

[0338] The partition information of a CU can indicate whether the CU is partitioned. The partition information can be a 1-bit flag. All CUs except SCUs can include partition information. For example, the partition information of a CU that is not further partitioned can be the first value, '0', and the partition information of a CU that is being partitioned can be the second value, '1'.

[0339] Quad Tree (QT) partitioning may mean that one CU is partitioned into four CUs. When a parent CU is partitioned into four child CUs, the width and height of each child CU may be half the width and half the height of the parent CU, respectively.

[0340] A binary tree (BT) split may mean that one CU is split into two CUs. For example, if a parent CU is split into two child CUs, the width or height of each child CU may be half the width or half the height of the parent CU.

[0341] A Ternary Tree (TT) partition may mean that a single CU is partitioned into three CUs. For example, if a parent CU is partitioned into three child CUs, the three child CUs can be created by partitioning the width or height of the parent CU in a ratio of 1:2:1. The width or height of the child CUs may be 1 / 4, 1 / 2, and 1 / 4 of the width or height of the parent CU, respectively.

[0342] In Fig. 2, QT type segmentation was applied to the first CTU. QT segmentation, BT segmentation, and TT segmentation were applied to the second CTU.

[0343] To partition a CTU, at least one of different types of partitions, such as QT partitioning, BT partitioning, and TT partitioning, may be applied to the CTU. Different types of partitions may be applied based on specific priorities.

[0344] For example, QT partitioning may be preferentially applied to a CTU. A CU to which QT partitioning can no longer be applied may correspond to a leaf node of QT. A CU that is a leaf node of QT may be a root node of BT and / or TT. A CU that is a leaf node of QT may be partitioned into a BT or TT form, or may not be partitioned any further. In this case, QT partitioning may not be applied again to a CU that is created by applying a BT or TT partition to a CU that is a leaf node of QT.

[0345] The partitioning of a CU corresponding to each node of QT can be signaled using QT partitioning information. The QT partitioning information can be a flag. The QT partitioning information of a unit can be information indicating whether the unit is partitioned in a QT form. A first value of the QT partitioning information, '0', can indicate that the CU is not partitioned in a QT form. The QT partitioning information having a first value can indicate a multi-type tree (MTT) partitioning. The MTT partitioning can include a BT partitioning and a TT partitioning. A second value of the QT partitioning information, '1', can indicate that the CU is partitioned in a QT form.

[0346] There may be no priority between BT and TT splits. That is, a CU corresponding to a leaf node of QT may be split into either BT or TT forms. Furthermore, a CU generated by BT or TT splits may be split again into BT or TT forms, or may not be split any further.

[0347] A CU corresponding to a leaf node of QT can become the root node of MTT. For each CU corresponding to an MTT node, the CU may further include MTT-type split direction information and split type information.

[0348] Split direction information can indicate the split direction of MTT splitting. The first value of the split direction information, '0', can indicate that the CU is split horizontally. The second value of the split direction information, '1', can indicate that the CU is split vertically.

[0349] The partition type information can indicate the partition type used for multi-type tree partitioning. The first value of the partition type information, '0', can indicate that the CU is partitioned in the TT form. The second value of the partition type information, '1', can indicate that the CU is partitioned in the BT form.

[0350] Here, each of the aforementioned split direction information and split type information may be a flag having a specific length (e.g., 1 bit).

[0351] The CU's partition information may also include QT partition information, partition direction information, and partition shape information.

[0352] CUs that are no longer split by QT splitting, BT splitting, and TT splitting can be used as units for specific processing, such as prediction, transformation, quantization, inverse quantization, inverse transform, entropy encoding, and entropy decoding. That is, for specific processing, CUs may no longer be split. Therefore, splitting information for splitting such CUs into PUs and / or TUs, etc., may not exist in the bitstream.

[0353] On the other hand, if the size of a CU is larger than the maximum TU size, the CU can be recursively split until the size of the CU becomes smaller than or equal to the maximum TU size. For example, if the size of a CU is 64x64 and the maximum TU size is 32x32, the CU can be split into four 32x32 TUs for transformation. For example, if the size of a CU is 32x64 and the maximum TU size is 32x32, the CU can be split into two 32x32 TUs for transformation.

[0354] In such cases, information regarding whether a CU is split for transformation may not be separately signaled. Whether a CU is split may be determined by comparing the size of the CU (width / height) with the maximum TU size (width / height), without signaling. For example, if the width of the CU is greater than the width of the maximum TU size, the CU may be split into two vertically. Additionally, if the height of the CU is greater than the height of the maximum TU size, the CU may be split into two horizontally.

[0355] For example, the minimum size of a CU may be 4x4. For example, the maximum size of a transform block may be 64x64. For example, the minimum size of a transform block may be 4x4. The QT minimum size may be the minimum size of a CU corresponding to a leaf node of the QT. The MTT maximum depth may be the maximum depth of the path from the root node to the leaf node of the MTT.

[0356] 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 to the minimum size of the CU.

[0357] If the depth within the MTT of a CU corresponding to a node of the MTT is equal to the maximum depth of the MTT, the CU may not be split into BT shape and / or TT shape.

[0358] Based on the various sizes and depths of the CUs described above, each piece of information described in the embodiments may or may not be present in the bitstream.

[0359] Information about the maximum or minimum size described in the embodiments may be signaled at a higher level of the CU. In the embodiments, the higher 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.

[0360] The information described in the embodiments may be signaled separately for different types of slices. The different types of slices may include intra-slices and inter-slices.

[0361]

[0362] Processing blocks according to their properties

[0363] Whether a specific process described in the embodiments is applied / performed may be determined based on the properties of a block related to the specific process. Whether a specific process described in the embodiments is applied / performed may be determined based on whether the properties of a block related to the specific process satisfy a specific condition. For example, a block may include a target block, a neighboring 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.

[0364] The blocks to which the specific processing described in the examples is applied may have a square shape or a non-square shape.

[0365] In one embodiment, the block's attributes may include the block's size. Certain processing described in the embodiments may be applied / performed when certain conditions regarding the block's size are met.

[0366] In one embodiment, the specific conditions may include a minimum block size condition and a maximum block size condition. The blocks to which the minimum block size condition applies and the blocks to which the maximum block size condition applies may be different.

[0367] In one embodiment, a minimum block size and / or a maximum block size for a particular process may be predefined.

[0368] In one embodiment, the processing of the embodiment may be applied / performed when the size of the block is greater than or equal to the minimum block size and / or when the size of the block is less than or equal to the maximum block size. Alternatively, in one embodiment, the processing of the embodiment may be applied / performed when the size of the block is greater than the minimum block size and / or when the size of the block is less than the maximum block size.

[0369] 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 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 or equal to the minimum block size and less than or equal to the maximum block size. 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.

[0370] In one embodiment, the processing of the embodiment may be applied / performed only when the block size is a predefined block size.

[0371] In embodiments, the size of a block may be determined in various ways. For example, the size of a block may refer to the width or height of the block. The size of a block may refer to both the width and height of the block. The size of a block may refer to the area of ​​the block. The size of a block may refer to 1) a result value of a known formula using the width and height of the block, 2) a result value of a formula of the embodiment, or 3) a statistical value.

[0372] Additionally, for the first size, the processing of the first embodiment among the embodiments may be applied / performed, and for the second size, the processing of the second embodiment among the embodiments may be applied / performed.

[0373] In embodiments, the block size may be 2x2, 4x4, 8x8, 16x16, 32x32, 64x64 or 128x128, etc. Alternatively, in embodiments, the block size may be (2*SIZE X )x(2*SIZE Y ) etc. SIZE X can be one of the integers greater than or equal to 1. SIZE Y can be one of the integers greater than or equal to 1.

[0374]

[0375] Predictive information for prediction

[0376] Prediction information can be used to generate a prediction block for the target block.

[0377] The encoding device (110) can generate prediction information required for prediction and can generate a bitstream including the prediction information. The prediction information can be signaled from the encoding device (110) to the decoding device (150) via the bitstream. The decoding device (150) can obtain the prediction information from the bitstream and perform prediction on the target block using the prediction information, thereby generating a prediction block.

[0378] Prediction information may include intra-prediction information, inter-prediction information, and IBC prediction information. In 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 embodiments. Inter-prediction information may include information used for inter-prediction as described in embodiments. IBC information may include information used for IBC prediction as described in embodiments.

[0379]

[0380] Intra prediction

[0381] Figure 3 illustrates the structure of intra prediction according to one embodiment.

[0382] Intra prediction can be performed using reference samples and coding parameters of the target block. The reference sample can be a (restored) sample within the (restored) reference block. Alternatively, an intermediate prediction sample can be generated using a sample described in the embodiment, such as a reconstructed sample, and a reference sample can be generated again using the intermediate prediction sample. Processing described in the embodiment, such as filtering, can be applied when generating the reference sample.

[0383] A reference block may be a (spatial) neighboring block of the target block. The coding parameters may be coding parameters for the target block and / or coding parameters for the reference block. In intra prediction, a reference sample may mean a neighboring sample.

[0384] A prediction block can be generated by performing intra prediction on a target block according to an intra prediction mode based on reference samples within a target image and information related to the reference samples. The size of the target block and the size of the prediction block can be the same.

[0385] In embodiments, the prediction block may be a PU. Alternatively, the prediction block may correspond to a CU or TU described in the embodiments. The prediction block may have a square or rectangular shape.

[0386] An intra prediction mode can be expressed 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 the 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. The intra prediction mode for the target block may be selected from among directional modes and non-directional modes.

[0387] In the lower right rectangle representing the target block, the number '0' may represent the planar mode, which is a non-directional intra prediction mode. The number '1' may represent the DC mode, which is a non-directional intra prediction mode. In the lower right rectangle representing the target block, the arrows from the center to the periphery of the rectangle may represent the prediction directions of the directional intra prediction modes. In addition, the number indicated close to the arrow may represent an example of the mode value assigned to the intra prediction mode or the prediction direction of the intra prediction mode.

[0388] Intra prediction can be performed based on an intra prediction mode for the target block. One of the available intra prediction modes for the target block can be used as the intra prediction mode for the target block.

[0389] The number of intra prediction modes available to a target block may be a predefined value. Alternatively, the number of intra prediction modes available to a target block may be determined based on the properties of the prediction block. For example, the properties of the prediction block may include coding parameters such as shape, size, and color components.

[0390] For example, in Figure 3, the directional modes depicted by the dotted lines (i.e., the directional modes numbered between -14 and -1, or between 67 and 80) can only be applied to predictions for non-square blocks. Therefore, the number of intra prediction modes available for predictions for square blocks can be 67 (planar mode, DC mode, and 65 directional modes).

[0391] For example, the number of available intra prediction modes may vary depending on whether the color component of the block is a luma signal or a chroma signal. The number of available intra prediction modes for a block containing a luma component may be greater than the number of available intra prediction modes for a block containing a chroma component.

[0392] Intra prediction modes may include horizontal-below mode, horizontal mode, vertical mode, and vertical-right mode. The horizontal-below mode may be an intra prediction mode located below 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 whose mode values ​​are one 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 whose mode value is one of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, and 17 can be horizontal bottom modes.

[0393] The number of intra prediction modes and the mode number of each intra prediction mode described above may be merely exemplary. The number of intra prediction modes and the mode number of each intra prediction mode described above may be defined differently depending on the embodiment, implementation, and / or needs.

[0394] When the intra prediction mode is the planar mode, when generating a prediction block of a target block, a sample value of the prediction sample can be generated using a weighted sum (weighted sum) of an upper reference sample of the target sample, a left reference sample of the target sample, an upper right reference sample of the target block, and a lower left reference sample of the target block, depending on the position of the prediction sample within the prediction block.

[0395] When the intra prediction mode is DC mode, a prediction block may be generated based on an average of sample values ​​of a plurality of reference samples. The plurality of reference samples may include upper reference samples and left reference samples of the target block. The value of the prediction sample of the prediction block may be determined based on an average of the sample values ​​of the plurality of reference samples. In addition, filtering using the values ​​of the reference samples may be performed for specific rows and / or specific columns within the target block. The specific rows may be one or more upper rows adjacent to the upper reference samples. The specific columns may be one or more left columns adjacent to the left reference samples.

[0396] When the intra prediction mode is a directional mode, a prediction block can be generated using the top reference sample, the left reference sample, the top right reference sample, and / or the bottom left reference sample of the target block.

[0397] The intra prediction mode of the target block may 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 may be signaled.

[0398] For example, if the intra prediction modes of the target block and the neighboring block are the same, an indicator indicating that the intra prediction modes of the target block and the neighboring block are the same can be signaled.

[0399] For example, an indicator may be signaled that indicates an intra prediction mode that is the same as the intra prediction mode of the target block among the intra prediction modes of multiple neighboring blocks.

[0400] For example, if the intra prediction modes of the target block and neighboring blocks are different, 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.

[0401] Reference samples used for intra prediction for the target block may include lower left reference samples, left reference samples, upper left reference samples, upper reference samples, and upper right reference samples.

[0402] For example, the left reference samples may be reconstructed reference samples adjacent to the left side of the target block. The top reference samples may be reconstructed reference samples adjacent to the top side of the target block. The top left reference sample may be a reconstructed reference sample diagonally adjacent to the top left side of the target block. The 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 the left reference samples. The top right reference samples may be reference samples located on the right side of the top reference samples among samples located on the same line as the top sample line composed of the top reference samples.

[0403] 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 can be used to determine the sample value of a prediction sample of a prediction block. In FIG. 3, the direction of the intra prediction mode indicated by the arrow can represent the direction from the prediction sample to the reference sample. The direction of the intra prediction mode can 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 can be used as the sample value of at least one sample of the prediction block. Here, the specific reference sample and the at least one sample of the prediction block can be samples designated by a straight line in the direction of the intra prediction mode. In other words, the sample value of the specific reference sample can be copied as 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 can be the sample value of the reference sample located in the direction of the intra prediction mode based on the position of the prediction sample.

[0404] The 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 lines 0 to 3 may be used for intra prediction of the target block.

[0405] Each reference sample line of FIG. 3 may include one or more reference samples. A smaller number of a reference sample line may be a line of reference samples closer to the target block. Reference sample line 0 may be a line of reference samples immediately adjacent to the target block. When the upper 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 whose x-coordinate is X-1 or whose y-coordinate is Y-1. Here, the y-coordinates of the reference samples whose x-coordinate is X-1 may be Y-1 to Y+2H. The x-coordinates of the reference samples whose y-coordinate is Y-1 may be X-1 to X+2W. The reference samples of reference sample line A may be samples whose x-coordinate is XA-1 or whose y-coordinate is YA-1. Here, the y-coordinates of the reference samples whose x-coordinate is XA-1 may be YA-1 to Y+2H+A. The x-coordinates of reference samples whose y-coordinate is YA-1 can be XA-1 to X+2W+A. A can be 1, 2, or 3.

[0406] Instead of obtaining samples from the reconstructed neighboring blocks, samples of segment A and segment F can be derived using padding using the nearest samples from segment B and segment E, respectively.

[0407] A 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 of one of 0 to 3. The reference sample line index may be signaled.

[0408] When inter-color component intra prediction is used for a target block, a prediction block of a second color component can be generated based on a reconstructed block of a first color component for the target block. For example, the first color component can be a luma component, and the second color component can be a chroma component.

[0409] For intra prediction between color components, parameters between the first color component and the second color component can be derived based on a template. For example, the parameters can be parameters of a linear model.

[0410] 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.

[0411] Once the parameters are derived, a prediction block of a second color component for the target block can be generated by applying the reconstructed 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 / downsampling can be performed on the surrounding samples of the reconstructed block of the first color component and the reconstructed block of the first color component. When subsampling is performed, the derivation of the parameters and intra prediction between color components can be performed using the corresponding samples derived by the subsampling.

[0412] Intra Sub-Partitions (ISP) prediction may refer to sequential intra prediction for multiple sub-blocks generated by dividing a target block. In ISP prediction, a target block may be divided into two or four sub-blocks in the horizontal and / or vertical directions. The divided sub-blocks may be sequentially reconstructed. As intra prediction is performed on a sub-block, a sub-prediction block for the sub-block may be generated. Additionally, as inverse quantization and / or inverse transformation is performed on the sub-block, a sub-residual block for the sub-block may be generated. A reconstructed sub-block may be generated by adding the sub-prediction block to the sub-residual block. The reconstructed sub-block may be used as a reference sample for intra predictions for other sub-blocks to be processed subsequently.

[0413] In performing prediction on a target block, it can be determined whether samples included in a reconstructed neighboring block can be used as reference samples of the target block. If there is an unavailable sample among the samples of the neighboring block that cannot be used as a reference sample of the target block, a value generated by copying and / or interpolating using the sample value of at least one sample among the samples included in the reconstructed neighboring block can replace the sample value of the unavailable sample. If the value generated by copying and / or interpolating is replaced with the sample value of the sample, the sample can be used as a reference sample of the target block.

[0414] In intra prediction, the sample value of a prediction sample of a prediction block can be determined by the sample value of a reference sample. The position of the reference sample can be specified by the position of the prediction sample and the direction of the intra prediction mode. If the position specified by the position of the prediction sample and the direction of the intra prediction mode is an integer position, the sample value of one reference sample pointed to by the integer position can be used to determine the sample value of the prediction sample of the prediction block. If the position specified by the position of the prediction sample and the direction of the intra prediction mode is not an integer position, an interpolated reference sample can be generated based on two reference samples closest to the specified position. The sample value of the interpolated reference sample can be used to determine the sample value of the prediction sample. That is, when the position specified by the position of the prediction sample and the direction of the intra prediction mode indicates a gap between two reference samples, an interpolated sample value can be generated based on the sample values ​​of the two samples.

[0415]

[0416] Inter prediction

[0417] Figure 4 shows the structure of inter prediction to explain the inter prediction process according to one embodiment.

[0418] The rectangle illustrated in Fig. 4 can represent an image. Additionally, the arrow in Fig. 4 can represent a prediction direction.

[0419] Each picture composing a video can be classified into an I picture (i.e., an intra picture), a P picture (i.e., a uni-prediction picture), and a B picture (i.e., a bi-prediction picture) according to its coding type. Coding can be performed for each picture according to its coding type.

[0420] 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 referring to other images. For example, coding for the I-picture can be performed using intra prediction and / or IBC prediction.

[0421] Coding for P pictures and B pictures can be performed by at least one of intra prediction, IBC prediction, and inter prediction using a reference picture.

[0422] If the target picture is a P picture, coding for the target picture can be performed using unidirectional inter prediction using one reference picture list.

[0423] When the target picture is a B picture, coding for the target picture can be performed using unidirectional inter prediction or bidirectional inter prediction using two reference picture lists.

[0424] Below, inter prediction for a target block in inter mode according to an embodiment is specifically described.

[0425] 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 split prediction block.

[0426] 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.

[0427] Motion information can be derived using coding parameters, etc. For example, motion information can be derived using motion information of a reconstructed neighboring block, motion information of a call block, and / or motion information of a block adjacent to a call block.

[0428] In embodiments, a candidate list may be used for inter prediction. The candidate list may include multiple candidates. An index indicating a candidate used for inter prediction for a target block among the candidates in the candidate list 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. Furthermore, in order to specify a candidate by index, the order of the candidates within the candidate list may need to be consistent.

[0429] In one embodiment, prediction of 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. The motion information of the spatial candidate may be referred to as spatial motion information. The motion information of the temporal candidate may be referred to as temporal motion information.

[0430] A spatial candidate may be a restored spatial neighboring block that is spatially adjacent to the target block.

[0431] A spatial candidate may be a block that 1) exists within the target image, 2) has already been restored through decryption, and 3) is adjacent to the target block.

[0432] Spatial candidates may include the left block, the top block, the bottom left block, the top right block, and the top left block of the target block.

[0433] A temporal candidate may be a restored temporal neighboring block corresponding to a target block in a restored COL image.

[0434] In embodiments, the motion information of a spatial candidate may be motion information of a block containing the spatial candidate. The motion information of a temporal candidate may be motion information of a block containing the temporal candidate.

[0435] In inter prediction, a call (COL) block for a target block can be identified. The area of ​​the target block within the target image and the area of ​​the call block within the call image may be identical. In other words, a call block may be a block occupying a specific area within the call image. The specific area may correspond to the area of ​​the target block within the call image.

[0436] A temporal candidate may be a location inside and / or outside a call block within a call image.

[0437] For example, a call block may include a first call block and a second call block. When the upper left coordinates of a call block are (xP, yP) and the size of the call block is (nPSW, nPSH), the first call block may be a block occupying coordinates (xP + nPSW, yP + nPSH). The second call block may be a block occupying coordinates (xP + (nPSW >> 1), yP + (nPSH >> 1)). The second call block may be optionally used as a call block when the first call block is unavailable.

[0438] 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 as the predicted MV. Alternatively, the MV of the temporal candidate stored in the candidate list associated with inter prediction can be a scaled MV.

[0439] 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 and the target image of the target block. The second temporal distance may be the distance between the reference image and the call image of the call block.

[0440] The method by which motion information is derived can be determined by the inter prediction mode of the target block. For example, as the inter prediction mode, AMVP mode, merge mode, skip mode, merge mode with MVD, subblock merge mode, GPM, Combined Inter Intra Prediction (CIIP) mode, and affine inter mode can be used. In the embodiments below, each of the inter prediction modes is described.

[0441]

[0442] AMVP mode

[0443] When the AMVP mode is used as a prediction mode, an MV candidate list including one or more MV candidates can be generated using the MV of the spatial candidate, the MV of the temporal candidate, the history-based MV candidate, and the zero vector. At least one of the MV of the spatial candidate, the MV of the temporal candidate, and the zero vector can be determined and used as an MV candidate.

[0444] A spatial candidate may include a reconstructed spatial neighboring block. The MV of the reconstructed spatial neighboring block may be referred to as a spatial MV candidate (spatial motion vector candidate). A temporal candidate may include a called block and a block adjacent to the called block. The MV of the called block or the MV of a block adjacent to the called block may be referred to as a temporal MV candidate (temporal motion vector candidate). A history-based MV candidate may be an MV in a list including MVs of other blocks that were previously encoded / decoded prior to encoding / decoding of the target block.

[0445] The encoding device (110) can use the MV candidate list to determine an MV to be used for encoding the target block within the search range. The maximum number of MV candidates in the MV candidate list can be predefined. N can represent a predefined maximum number. For example, N can be 2. Alternatively, the maximum number of candidates can 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 a prediction MV of the target block among the MV candidates in the MV candidate list. The MV to be used for encoding the target block can be an MV that can be encoded at the minimum cost. The encoding device (110) can determine whether to use the AMVP mode in encoding the target block, and can generate AMVP mode usage information indicating whether the AMVP mode is used.

[0446] 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 a residual block. Inter prediction information may be signaled from an encoding device (110) to a decoding device (150) in the form of a bitstream.

[0447] 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 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, an MV candidate indicated by the MV candidate index can be selected as the prediction MV of the target block.

[0448] The MVD may represent the difference between the MV to be actually 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 to be actually used for inter prediction of the target block in order to use an MVD with the smallest possible size. The decoding device (150) may derive the MV of the target block by combining the MVD and the predicted MV. In other words, the MV of the target block derived by the decoding device (150) may be the sum of the MVD and the predicted MV candidate.

[0449] 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.

[0450] Meanwhile, the encoding device (110) can calculate the MVD based on the affine model. The affine control point MV of the target block can be derived based on the sum of the affine control point MV candidate and the MVD. The MV of each subblock within the target block can be derived using the affine control point MV.

[0451]

[0452] Merge mode

[0453] When merge mode is used, a merge candidate list including multiple merge candidates can be generated using motion information of spatial candidates and motion information of temporal candidates. The motion information can include 1) MV, 2) reference image index, and 3) reference direction. The merge candidate can be motion information.

[0454] Merge candidates may include 1) a spatial merge candidate generated based on a spatial candidate, 2) a temporal merge candidate generated based on a temporal candidate, 3) a history-based merge candidate, 4) an average merge candidate, and 5) a zero merge candidate.

[0455] A history-based merge candidate may be motion information within a list that includes motion information of other blocks that were previously encoded / decoded prior to encoding / decoding of the target block.

[0456] An average merge candidate may be a merge candidate generated based on the average of two merge candidates in the merge candidate list.

[0457] A zero merge candidate may be zero vector motion information. Zero vector motion information may be motion information whose MV is a zero vector.

[0458] 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 constructed in the encoding device (110) and the decoding device (150) through the predefined method and the predefined order.

[0459] 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.

[0460] Inter prediction information may include 1) merge mode usage information, 2) merge index, and 3) correction information, and may include a residual block. Inter prediction information may be signaled from an encoding device (110) in bitstream form to a decoding device (150) in bitstream form.

[0461] 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 information related to the merge mode, such as a merge index, from the bitstream.

[0462] The encoding device (110) can select an optimal merge candidate from among the merge candidates included in the merge candidate list, and set the value of the merge index to point to the selected merge candidate.

[0463] Correction information may be information used for correcting an MV. The encoding device (110) may generate the correction information. The decoding device (150) may perform correction on the MV of the merge candidate selected by the merge index based on the correction information, thereby deriving a corrected MV. The corrected MV may be used as the MV of the target block.

[0464] 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 size information. The correction usage information may indicate whether correction is used 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 with an MVD.

[0465] In merge mode, prediction for a target block can be performed using a merge candidate pointed to by a merge index among the merge candidates included in the merge candidate list.

[0466] Motion 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.

[0467] In one embodiment, the merge candidates in the merge candidate list may be specific modes that derive inter-prediction information. The merge candidate may be information indicating a specific mode that derives inter-prediction information. Inter-prediction information of the target block may be derived according to the specific mode indicated by the merge candidate. From this perspective, a specific mode may be considered a specific inter-prediction information derivation mode or a specific motion information derivation mode. A specific mode may include a series of processes that derive inter-prediction information.

[0468] Inter prediction information of a target block can be derived based on a specific mode indicated by a merge candidate selected by a merge index among the merge candidates in the merge candidate list. For example, the specific modes may include a subblock-level motion information derivation mode and an affine motion information derivation mode, and may include other modes for deriving motion information described in the embodiments.

[0469] Skip mode may be a mode that does not use residual blocks. That is, when skip mode is used, the reconstructed block may be identical to the predicted block. The description of 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, 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.

[0470] The subblock merge mode may be a mode in which motion information of a target subblock is derived 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 a call subblock of the target subblock.

[0471] In GPM, a first prediction block and a second prediction block can be generated using two pieces of motion information for a target block. For each coordinate of the target block, a final prediction sample of a final prediction block can be generated using a weighted sum of the first prediction sample of the first prediction block and the second prediction sample of the second prediction block.

[0472] Here, the first weight for the weighted consensus first prediction sample and the second weight for the weighted consensus second prediction sample can be determined based on the boundary of the GPM. The boundary can represent a dividing line that divides the target block. Based on the boundary, the target block can be divided into a first divided region and a second divided region.

[0473] 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 a 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.

[0474] Combined Inter-Intra Prediction (CIIP) mode may be a mode that derives a prediction sample of a target block using a weighted sum of prediction samples generated by inter prediction and prediction samples generated by intra prediction.

[0475] In the aforementioned modes, self-improvement of the derived motion information can be performed, and the improved motion information can be used as motion information for the target block. For example, blocks within a specific region determined based on the derived motion information can be searched, and the motion information of the block with the smallest sum of absolute differences (SAD) value among the searched blocks can be used as the improved motion information for the target block. The specific region can be a square region within a reference image specified by the motion information. The point indicated by the motion information can be the center of the specific region.

[0476] In the aforementioned modes, compensation for prediction samples derived through inter prediction can be performed using optical flow.

[0477]

[0478] Figure 5 shows the order in which spatial candidates are added to the candidate list according to one embodiment.

[0479] In Fig. 5, the locations of spatial candidates are shown.

[0480] The large block in the center can represent the target block. The five smaller blocks adjacent to the target block can represent spatial candidates.

[0481] The coordinates of the target block can be (xP, yP), and the size of the target block can be (nPSW, nPSH).

[0482] A spatial candidate A0 may be a block adjacent to the lower left of the target block. A0 may be a block that occupies samples at coordinates (xP - 1, yP + nPSH).

[0483] A 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. Alternatively, A1 may be a block adjacent to the top of A0. A1 may be a block that occupies a sample at coordinates (xP - 1, yP + nPSH - 1).

[0484] A spatial candidate B0 may be a block adjacent to the upper right of the target block. B0 may be a block that occupies a sample at coordinates (xP + nPSW, yP - 1).

[0485] A 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. Alternatively, B1 may be a block adjacent to the left of B0. B1 may be a block that occupies a sample at coordinates (xP + nPSW - 1, yP - 1).

[0486] A spatial candidate B2 may be a block adjacent to the upper left of the target block. B2 may be a block that occupies a sample at coordinates (xP - 1, yP - 1).

[0487] As shown in Figure 5, when adding spatial candidates to the candidate list, B1, A1, The order of B0, A0 and B2 can be used, i.e. B1, A1, Available spatial candidates can be added to the candidate list in the order of B0, A0, and B2. The order in which the spatial candidates are added to the merge candidate list illustrated in Fig. 5 may be merely an example.

[0488] 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.

[0489] To include a spatial or temporal candidate in the candidate list, its availability can be determined. If the candidate block is outside the boundaries of an image, slice, or tile, the candidate block's availability can be set to false. The phrase "availability is set to false" can mean "it is set to non-availability."

[0490] The maximum number of candidates in a 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 default value of N can be 5.

[0491]

[0492] IBC mode

[0493] IBC mode may be an intra-block copy prediction mode that generates prediction blocks for target blocks by referencing already-restored regions within the target image. In this respect, IBC mode may also be referred to as a current image reference mode. A block vector (BV) may be used to specify the already-restored region.

[0494] Whether a 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.

[0495] In IBC mode, a prediction block of a target block can be generated based on a block vector (BV). The BV can specify a reference block. The BV can indicate 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 in the embodiments can also be applied to the BV.

[0496] The IBC mode may include skip mode, merge mode, and AMVP mode. The description of the AMVP mode, merge mode, and skip mode of the embodiments may also be similarly applied to the AMVP mode, merge mode, and skip mode of the IBC mode.

[0497] In skip mode or merge mode, a merge candidate list can be constructed, and a merge index can specify one merge candidate among the merge candidates in the merge candidate list. The BV of the specified merge candidate can be used as the BV of the target block.

[0498] In AMVP mode, BVD can be used. The description of MVD in the embodiments can also be applied to BVD.

[0499] The reference block in IBC mode may be limited to a block within an already reconstructed 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 limited 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 that contains the target block. The specific size may be 64x64.

[0500]

[0501] Transformation and quantization

[0502] A quantized level can be generated by performing transformation and / or quantization on a residual block. The residual block can represent the difference between the original block and the predicted block. A reconstructed residual block can be generated by performing inverse quantization and / or inverse transformation on the quantized level. The reconstructed residual block can represent the difference between the reconstructed block and the predicted block.

[0503] When a transformation or inverse transformation is performed, a separable transformation or a 2-dimensional (2D) non-separable transformation can be performed on the residual block. A separable transformation can be a transformation that performs 1-dimensional (1D) transformations on the residual block in each of the horizontal and vertical directions.

[0504] The transform kernel used for the transformation may include various DCT kernels such as DCT type 2 (DCT-II), 1) DST kernels, and 3) kernels induced by training. For 1D transform, the DCT type and DST type may include DCT-V, DCT-VIII, DST-I, and DST-VII in addition to DCT-II.

[0505] A set of transforms may be used to determine the DCT type, DST type, or learning-induced kernel to be used for the transformation. Each transform set may include multiple transform candidates. Each transform candidate may be a DCT type, a DST type, or a learning-induced kernel.

[0506] The encoding device (110) can perform transformation and inverse transformation using transformation candidates included in the transformation set. The decoding device (150) can perform inverse transformation using transformation candidates included in the transformation set. Transform selection information indicating which transformation candidate among a plurality of transformation candidates included in the transformation set applied to the residual block is used can be signaled. The transformation selection information can include vertical transformation selection information and horizontal transformation selection information. The vertical transformation selection information can indicate which transformation among the transformations included in the transformation set is used for vertical transformation. The horizontal transformation selection information can indicate which transformation among the transformations included in the transformation set is used for horizontal transformation.

[0507] The transform may include at least one of a primary transform and a secondary transform. A primary transform coefficient may be generated by performing a primary transform on a residual block, and a secondary transform coefficient may be generated by performing a secondary transform on the transform coefficient. Here, the transform coefficient may include a primary transform coefficient and a secondary transform coefficient.

[0508] The primary transformation may mean Multiple Transform Selection (MTS), which applies different transformations for each of the 1D directions (i.e., vertical and horizontal directions).

[0509] A secondary transform may be a transform for improving the energy concentration of the transform coefficients generated by the primary transform. The secondary transform may be 1) a separable transform like the primary transform, or 2) a 2D non-separable transform. The 2D non-separable transform may refer to a low frequency non-separable transform (LFNST) or a non-separable primary transform (NSPT).

[0510] NSPT can be applied to specific block sizes such as 4x4, 4x8, 8x4, 4x16, 16x4, 8x8, 8x16, and 16x8 for intra coding.

[0511] The primary transform can be performed using at least one of a plurality of predefined transform methods. For example, the plurality of predefined transform methods can include DCT, DST, and KLT. In addition, the primary transform can be a transform having various transform types according to a transform kernel function defining DCT and DST. For example, the primary transform can include a plurality of transforms 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 transform kernels.

[0512] In one embodiment, the transform type may be determined based on coding parameters associated with the target block. For example, the transform type may be determined based on one or more of: 1) a prediction mode of the target block (e.g., one of intra prediction and inter prediction), 2) a size of the target block, 3) a shape of the target block, 4) an intra prediction mode of the target block, 5) a component of the target block (e.g., one of a luma component and a chroma component), and 6) a split type applied to the target block (e.g., one of QT, BT, TT, and non-split).

[0513] As with the first-order transformation, a set of transformations can also be defined for the second-order transformation. The methods for deriving and / or determining the set of transformations of the embodiments can be applied to both the first-order transformation and the second-order transformation.

[0514] In one embodiment, a primary transformation and / or a secondary 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 primary transformation and / or the secondary transformation is applied.

[0515] For example, a first-order transform and / or a second-order transform may be applied to one or more of the signal components, including the luma component and the chroma component.

[0516] In one embodiment, the transform selection information may include primary transform usage information and secondary transform usage information. The primary transform usage information may indicate whether the primary transform is applied to the residual block of the target block. The secondary transform usage information may indicate whether the secondary transform is applied to the residual block of the target block.

[0517] In one embodiment, whether a primary transform and / or a secondary transform is applied may be determined based on coding parameters for the target / neighboring blocks, such as the size and shape of the target / neighboring blocks.

[0518] In one embodiment, the transform selection information may include primary transform selection information and secondary transform selection information. The primary transform selection information may indicate a transform method to be applied to a residual block among a plurality of transform methods that may be used in the primary transform. The primary transform selection information may be a primary transform index. The secondary transform selection information may indicate a transform method to be applied to a transform coefficient among a plurality of transform methods that may be used in the secondary transform. The secondary transform selection information may be a secondary transform index.

[0519] In one embodiment, the transformation methods of the first and second transformations may each be derived based on specific information such as coding parameters. For example, the coding parameters may include coding parameters for target / neighboring blocks.

[0520] In embodiments, information related to transformation, such as transformation selection information, and sub-information of the transformation selection information may be signaled for a specific target. For example, the specific target may be a CU.

[0521] Information related to transformation, 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.

[0522] Quantized levels can be generated by performing quantization on the result or residual block generated by performing the first transform and / or the second transform.

[0523] The description of the transformation described above can also be applied to the inverse transformation. In this application, the reverse processing of the processing described for the transformation can be performed in the inverse transformation. The term "transformation" in the name related to the transformation can be changed to "inverse transformation." Furthermore, the input of the transformation can be considered the output of the inverse transformation. The output of the transformation can be considered the input of the inverse transformation. The decoding device (150) can obtain information related to the transformation, such as transformation selection information, and can perform the reverse processing of the processing related to the transformation indicated by the information related to the transformation using the information related to the transformation.

[0524] A target block may include multiple subblocks. Each subblock may be defined according to a minimum block size or a minimum block shape. The target block may be divided into multiple subblocks, and each subblock may include coefficients of sizes such as 4x4, 2x8, and 8x2. The target block may be a transform block. Transform coefficients or quantized levels may be expressed in the form of a block. The transform coefficients may be quantized transform coefficients.

[0525] The transform coefficients or quantized levels can be scanned according to at least one of the scanning types, such as diagonal scanning, vertical scanning, and horizontal scanning. The diagonal scanning can be right-upper diagonal scanning or left-lower diagonal scanning.

[0526] For example, coefficients can be transformed or arranged into a one-dimensional vector by scanning the coefficients of a block using diagonal scanning. Vertical scanning can scan coefficients in the form of two-dimensional blocks in the column direction. Horizontal scanning can scan coefficients in the form of two-dimensional blocks in the row direction.

[0527] The scanning type for coefficients can be determined based on coding parameters such as the intra prediction mode, block size, and block shape. For example, whether diagonal scanning, vertical scanning, or horizontal scanning is used can be determined based on coding parameters such as the intra prediction mode, block size, and block shape. A block can be a transform unit.

[0528] Scanning for each scanning type can start at a specific starting point and end at a specific ending point.

[0529] In scanning, a scanning order based on the scanning type may first be applied between subblocks. Next, a scanning order based on the scanning type may be applied to transform coefficients or quantized levels within the subblock.

[0530] The encoding device (110) can perform entropy encoding on transform coefficients or quantized levels to generate a bitstream including entropy-encoded transform coefficients or entropy-encoded quantized levels.

[0531] The decoding device (150) can obtain entropy-encoded transform coefficients or entropy-encoded quantized levels from a bitstream and perform entropy decoding to generate transform coefficients or quantized levels. The coefficients can be arranged in the form of two-dimensional blocks through inverse scanning. The arrangement of the inverse scanning can be a rearrangement opposite to the arrangement of the scanning.

[0532] Inverse scanning of coefficients can generate inversely scanned transform coefficients or inversely scanned quantized levels. At this time, the inverse scanning types of the inverse scanning can include diagonal scanning, vertical scanning, and horizontal scanning, and the inverse scanning type of the inverse transformation corresponding to the scanning type of the transformation can be selected.

[0533] In the decoding device (150), inverse quantization can be performed on (inversely scanned) coefficients. Depending on whether a second inverse transform is performed, a second inverse transform can be performed on the result generated by performing inverse quantization. In addition, depending on whether a first inverse transform is performed, a first inverse transform can be performed on the result generated by performing the second inverse transform. A restored residual block can be generated by selectively performing the second inverse transform and the first inverse transform on the coefficients.

[0534]

[0535] Filtering

[0536] To improve the image quality, filtering may be performed on blocks. The values ​​of target samples may be determined or updated through filtering.

[0537] 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 restored sample, and a restored sample with filtering applied.

[0538] The target sample may be a sample within one or more of a target picture, a target slice, a target CTB, a target block, a reference sample line, and a 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 transform block, a prediction block, a reference block, a residual block, and a reconstruction block.

[0539] In embodiments, the filtering process described as being applied to one object may also be applied to other objects. For example, the filtering process described in a specific in-loop filtering may also be applied to transform blocks, prediction blocks, reference blocks, and residual blocks.

[0540] A specific type of filtering may be used for the filters of the embodiments. The type of filtering may include filter taps (or filter tap lengths), filter shapes, filter strengths, filter coefficients (or weights), and offsets.

[0541] The filter tab may indicate the number of input samples used for the filter. The input samples may include the target sample. Alternatively, the input samples may include a specific value determined for the target sample. The input samples may include one or more reference samples. The one or more reference samples may be determined based on an attribute of the target block described in the embodiments. The attribute may include a coding parameter. For example, an attribute of the target sample may include a position of the target sample. One or more reference samples may be specified based on a relative position with respect to the position of the target sample.

[0542] A filter shape can represent the shape formed by input samples. A specific value determined for a target sample can be considered a target sample. In other words, if a specific value determined for a target sample is used as an input sample of a filter, the target sample can also be considered to form a filter shape.

[0543] The number of samples whose values ​​are determined by filtering may be multiple. The filter strength may indicate the range of samples whose values ​​are determined by filtering. The filter strength may be either a strong filtering strength or a weak filtering strength. The number of samples whose values ​​are determined by a strong filtering strength may be greater than the number of samples whose values ​​are determined by a weak filtering strength. Alternatively, the filter strength may indicate the range of values ​​that are changed by filtering. The range of sample values ​​that are changed by a strong filtering strength may be wider than the range of sample values ​​that are changed by a weak filtering strength.

[0544] The filter coefficients can be coefficients or weights of the input samples.

[0545] An offset can be a specific value that is added to the result calculated using the values ​​and coefficients of the input samples, such as a weighted sum.

[0546] Filtering, interpolation, and sampling may have in common that they update the values ​​of samples. Therefore, the description of any one of filtering, interpolation, and sampling in the embodiments may also apply to any other of filtering, interpolation, and sampling. Here, sampling may include at least one of upsampling, downsampling, and subsampling.

[0547] Filtering may include filtering performed by predictor (123) and predictor (163), etc.

[0548] In encoding a target block, a prediction error may exist between the original samples of the original block and the prediction samples of the prediction block. To reduce the prediction error, filtering may be performed on at least one of the prediction samples of the prediction block and the reference samples referenced for prediction.

[0549] For example, in intra prediction, the reference sample may include one or more of the upper left reference sample, the upper reference sample, the upper right reference sample, the left reference sample, and the lower left reference sample. Filtering on the predicted sample may be performed by applying specific weights to the predicted sample, the left reference sample, the upper reference sample, and / or the upper left reference sample, respectively.

[0550] Filtering of 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 the filtering is applied, the filtering weights, 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.

[0551] For example, the properties of the target block may include information related to the target block described in the embodiments, such as 1) size of the target block, 2) prediction mode, 3) intra prediction mode, 4) reference sample line, 5) sample value, and 6) coding parameter.

[0552] For example, the attributes of a prediction sample may include information related to the prediction sample described in the embodiments, such as 1) a sample value of the prediction sample and 2) a location within a target block, and may include coding parameters related to the prediction sample.

[0553] Filtering may include in-loop filtering performed by filter (130) and filter (170), etc.

[0554]

[0555] Figure 6 illustrates multiple in-loop filters according to an example.

[0556] The plurality of in-loop filters of the in-loop filtering may include one or more of Luma Mapping with Chroma Scaling (LMCS), a deblocking filter, a Sample Adaptive Offset (SAO), and an Adaptive Loop Filter (ALF).

[0557] Multiple in-loop filters can be connected sequentially. For example, the multiple in-loop filters can be connected in the order of LMCS, deblocking filter, SAO, and ALF. Furthermore, the multiple in-loop filters can be connected in any order among all available permutations of the multiple in-loop filters. The output from one of the multiple in-loop filters can be used as the input to the next filter.

[0558] As illustrated in FIG. 6, an input image may be input to the first filter. The input image may be a block described in the embodiments. For example, the input image may be a reconstructed 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 described in the embodiments. For example, the output image may be a filtered reconstructed image generated by a filter (130) or a filter (170).

[0559] The target block can represent an image input to the filter. The filtered target block can represent an image output from the filter.

[0560] LMCS may include luma signal mapping to a luma signal of a target block and chroma signal scaling to a chroma signal of the target block.

[0561] Luma signal mapping can perform codeword redistribution for the luma signal.

[0562] Luma signal mapping can include forward mapping and reverse mapping. In forward mapping, the existing dynamic range can 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 reverse mapping, reverse mapping is performed from the mapped dynamic range to the existing dynamic range.

[0563] Chroma scaling can correct chroma signals based on the correlation between a luma signal and a corresponding chroma signal.

[0564] Forward mapping can be performed between inter prediction for a luma signal and reconstruction for the luma signal, and between inter prediction for the luma signal and chroma scaling. Backward mapping can be performed between reconstruction for the luma signal and in-loop filtering for the luma signal. Chroma scaling can be performed between inverse transformation and reconstruction for the chroma signal.

[0565] According to this structure, inverse quantizations for luma and chroma signals, inverse transformations for luma and chroma signals, prediction for luma signals, and restoration for luma signals can be performed within the mapped dynamic range. In-loop filterings for luma and chroma signals, inter predictions for luma and chroma signals, intra prediction for chroma signals, and restoration for chroma signals can be performed within the existing dynamic range.

[0566] A deblocking filter can remove block distortion occurring at boundaries between blocks within a restored image. For example, the blocks may be transform blocks. Furthermore, 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.

[0567] Deblocking filters can be applied to vertical and horizontal boundaries between blocks. After filtering the vertical boundaries of blocks, filtering can be performed again on the horizontal boundaries of the filtered blocks.

[0568] A deblocking filter may be applied selectively. Whether to apply a deblocking filter to a target block may be determined based on at least one of the sample(s) contained within a specific number of columns or rows within the target block and the sample(s) contained within a specific number of columns or rows within a neighboring block adjacent to a specific boundary.

[0569] When a deblocking filter is applied to a target block, the filter to be applied may be determined based on the strength of the required deblocking filtering. In other words, among multiple other filters, a filter determined based on the strength of the deblocking filtering may be applied to the target block. The multiple filters may include one of a long-tap filter, a strong filter, a weak filter, and a Gaussian filter.

[0570] The maximum length of the deblocking filter can be determined based on the properties of the target block, such as the size of the target block, the components of the target block, and the coding parameters.

[0571] SAO can compensate for distortion between the original and reconstructed images on a sample-by-sample basis. To compensate, SAO can apply an appropriate offset to the sample values ​​of each sample. That is, the offset can be added to the sample values.

[0572] An offset can be determined for the target block. For example, an 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.

[0573] SAO may include SAO using Edge Offset (EO) and SAO using Band Offset (BO). Depending on the characteristics of samples within a specific block, such as a CTU, whether SAO using EO or SAO using BO may be performed may be determined.

[0574] In SAO using EO, distortion correction of samples can be performed based on the direction of the edge within the target block. Pattern classes of EO can include horizontal patterns, vertical patterns, 135 degree diagonal patterns, and 45 degree diagonal patterns. For a target block, information indicating a pattern class applied to the target block and multiple offsets of the pattern class can be signaled. There can be four offsets. For a target sample within the target block, adjacent samples of the target sample can be determined based on the direction of the pattern class. An offset to be applied to the target sample can be determined based on the pattern of the adjacent samples.

[0575] In an offset using BO, distortion of a sample can be corrected by classifying the brightness values ​​of samples within a target block into specific bands. The bit depth of an input image can be divided into m sections. For example, m can be 32. The specific bands can be n consecutive sections among the m sections. For example, n can be 4. N offsets for the n sections can be signaled. Additionally, information indicating a first section selected as one of the n sections among the m sections can be signaled. The offset of the section to which the target sample corresponds can be added to the sample value of the target sample of the target unit.

[0576] ALF can compensate for distortion between the restored image and the original image.

[0577] The filter coefficients of ALF can be signaled via the bitstream.

[0578] The filter shape of ALF can be determined by the components of the target block. For example, a 7x7 diamond-shaped filter can be used for the luma component. A 5x5 diamond-shaped filter can be used for the chroma component.

[0579] In ALF, the characteristics of a specific block can be determined for a specific block, and the class of the specific block can be determined based on the characteristics. In other words, the determination of characteristics and class of ALF can be performed in units of 4x4 blocks. Filter coefficients can be calculated based on the class. A specific block can be a 4x4 block.

[0580] 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. Rotation, vertical symmetry, and / or diagonal symmetry transformations can be applied to the filter based on the gradient of the specific block.

[0581] Information regarding whether ALF applies can be signaled for specific units, such as CTB.

[0582] An index indicating a filter to be applied to a specific unit among 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 identically predefined in the encoding device (110) and the decoding device (150). The filter coefficients of the filters configured using the parameter set may be determined based on coding parameters.

[0583]

[0584] Entropy encoding and entropy decoding

[0585] Figure 7 illustrates entropy encoding and entropy decoding according to an example.

[0586] The processes of entropy encoding by the entropy encoder (139) are illustrated at the top of Fig. 7.

[0587] The entropy encoder (139) may include a context modeler, a binarization unit, and an entropy encoder. The context modeler may include a context selection unit and a context memory.

[0588] The binarization unit can generate bins 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 bins.

[0589] Information about syntactic elements and bins can be provided from the binarization unit to the context selection unit.

[0590] A context modeler can perform context updates.

[0591] Context can mean occurrence probability information for each bin for syntactic elements that have already been encoded.

[0592] The context modeler can update the context to apply current probability information to the entropy encoding of the bins of the syntactic elements of the target block. The updated context can be stored in the context memory. At this time, the updated context corresponding to the syntactic elements of the target block (or bins within the syntactic elements of the target block) can be derived by the context modeler.

[0593] 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 the context memory and used as an updated context for entropy encoding of the bins of the syntactic element of the target block.

[0594] The updated context can be used for entropy encoding of syntactic elements of the target block.

[0595] The entropy encoding unit can generate encoded information about syntactic elements of a target block by performing entropy encoding using the generated bins and the updated context, and can generate a bitstream including the encoded information. The entropy encoding unit can use at least one of an arithmetic encoding method and a bypass encoding method.

[0596] The processes of entropy decryption by the entropy decoder (161) are shown at the bottom of Fig. 7.

[0597] The entropy decoder (161) may include a context modeler, an entropy decoder, and an inverse binarizer. The context modeler may include a context selection unit and a context memory.

[0598] A context modeler can perform context updates.

[0599] Context can mean the occurrence probability information of each bin for syntactic elements that have already been decoded.

[0600] The context modeler can update the context to apply the currently decoded probability information to entropy decoding for the bins of the syntactic elements of the target block. The updated context can be stored in the 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) can be derived by the context modeler.

[0601] The context selector can select a context corresponding to a blank of a syntactic element of a target block. The selected context can be loaded from the context memory and used as an updated context for entropy decoding of the syntactic element of the target block.

[0602] The updated context can be used for entropy decoding of syntactic elements of the target block.

[0603] The entropy decoding unit can generate bins for the delimiting 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 can use at least one of an arithmetic decoding method and a bypass decoding method.

[0604] The debinarization unit can obtain a syntactic element of the target block by performing debinarization on at least one of the generated bins. The debinarization may be a process of converting at least one of the bins into a form of a syntactic element.

[0605] Information about syntactic elements and bins can be provided from the de-binarization unit to the context selection unit.

[0606] A syntax element may be one of the coding parameters described in the embodiments.

[0607]

[0608] Methods for binarization, debinarization, entropy encoding, and entropy decoding

[0609] In embodiments, one or more of the binarization methods, inverse binarization methods, entropy encoding methods and entropy decoding methods listed below may be used to perform signaling for specific information.

[0610] - Signed 0-th order Exponential Golomb binarization / debinarization method (abbreviated as se(v))

[0611] - k-order exponential-Golomb binarization / inverse binarization method with sign (abbreviated as sek(v))

[0612] - 0-order exponent-Golomb binarization / inverse binarization method for unsigned positive integers (abbreviated as ue(v))

[0613] - k-order exponential-Golomb binarization / inverse binarization method for unsigned positive integers (abbreviated as uek(v))

[0614] - Fixed-length binarization / debinarization method (abbreviated as f(n))

[0615] - Truncated Rice binarization / debinarization method or truncated unary binarization / debinarization method (abbreviated as tu(v))

[0616] - Truncated binary binarization / debinarization method (abbreviated as tb(v))

[0617] - Context-adaptive arithmetic encoding / decoding method (abbreviated as ae(v))

[0618] - bit string in bytes (abbreviated as b(8))

[0619] - Signed integer binarization / debinarization method (abbreviated as i(n))

[0620] - Unsigned positive integer binarization / debinarization method (abbreviated as u(n)) ('u(n)' can also mean fixed-length binarization / debinarization method.)

[0621] - Unary binarization / inverse binarization method

[0622]

[0623] Image encoding using prediction

[0624] FIG. 8 is a flowchart of a method for predicting a target block and a method for generating a bitstream according to one embodiment.

[0625] The prediction method and bitstream generation method of the target block of the embodiment can be performed by the encoding device (110). The embodiment can be a part of the encoding method of the target block or the video encoding method.

[0626] At step (810), the processor (120) can derive a prediction mode for the target block.

[0627] The prediction mode may be one of the modes using prediction described in the embodiments.

[0628] The processor (120) can derive a prediction mode for the target block by considering rate-distortion costs for prediction modes among the prediction modes available for the target block.

[0629] For example, the prediction mode may be an intra prediction mode. The prediction for the target block may be an intra prediction mode, and the intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0630] For example, the processor (120) can derive an intra prediction mode for the target block by considering rate-distortion costs for intra prediction modes among the intra prediction modes available for the target block.

[0631] At step (820), the processor (120) can construct a reference sample for prediction.

[0632] The reference sample may be a reference sample used for prediction as described in the embodiments.

[0633] 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.

[0634] At step (830), the processor (120) can perform prediction on the target block using the derived prediction mode. The processor (120) can generate a prediction block for the target block through the prediction.

[0635] For example, in step (830), the processor (120) can perform intra prediction on the target block using the derived intra prediction mode. The processor (120) can generate a prediction block for the target block through intra prediction.

[0636] Information about an encoded target block can be generated by performing prediction on the target block using the derived prediction mode.

[0637] For example, information about an encoded target block can be generated by performing intra prediction on the target block using an induced intra prediction mode.

[0638] A prediction block can be generated by predicting a target block using an induced prediction mode, and a residual block, which is the difference between the target block and the predicted block, can be generated. By applying transformation and quantization to the residual block, information about the encoded target block can be generated.

[0639] For example, a predicted block can be generated by intra-prediction on a target block using an induced intra-prediction mode, and a residual block, which is the difference between the target block and the predicted block, can be generated. By applying transformation and quantization to the residual block, information about the encoded target block can be generated.

[0640] Information about an encoded target block may include transform and quantized coefficients for the target block. Information about an encoded target block may include coding parameters for the target block.

[0641] At step (840), the processor (120) can generate a bitstream.

[0642] The bitstream may contain information about the encoded target block.

[0643] The bitstream may contain prediction information.

[0644] The prediction information may be information regarding the prediction of the target block described above. The information regarding the prediction of the target block may include coding parameters related to the target block and / or sub-block, etc., for the prediction described in the embodiments.

[0645] For example, the prediction information may be information regarding intra prediction of the target block described above. The information regarding intra prediction of the target block may include coding parameters related to the target block and / or sub-block, etc., for intra prediction described in the embodiments.

[0646] The prediction information may be generated in step (840), or may be generated at least partially in steps (810, 820, and 830).

[0647] The processor (120) can store the generated bitstream in the storage (140). Alternatively, the communication device (149) can transmit the bitstream to the decryption device (150).

[0648] The processor (120) can perform entropy encoding on prediction information and generate a bitstream including entropy-encoded prediction information.

[0649] The embodiment can be combined with the operations of the encoding device (110) described above with reference to FIG. 1. For example, the operations of steps (810), (820), and (830) can be performed by the predictor (123). The operations of step (840) can be performed by the entropy encoder (139). Additionally, operations performed in other components of the encoding device (110) before, after, and between steps (810), (820), (830), and (840) can be performed.

[0650]

[0651] Image decoding using prediction

[0652] FIG. 9 is a flowchart of a method for predicting a target block using a bitstream according to one embodiment.

[0653] The method of predicting a target block using the bitstream of the embodiment can be performed by a decoding device (150). The embodiment can be a part of a decoding method of a target block or a video decoding method.

[0654] In step (910), the communication device (189) can obtain a bitstream. The communication device (189) can receive the bitstream from the encoding device (110).

[0655] The bitstream may contain information about the encoded target block.

[0656] Information about an encoded target block may include transform and quantized coefficients for the target block. Information about an encoded target block may include coding parameters for the target block.

[0657] The bitstream may contain prediction information.

[0658] The prediction information may be information regarding the prediction of the target block described above. The information regarding the prediction of the target block may include coding parameters related to the target block and / or sub-block, etc., for the prediction described in the embodiments.

[0659] For example, the prediction information may be information regarding intra prediction of the target block described above. The information regarding intra prediction of the target block may include coding parameters related to the target block and / or sub-block, etc., for intra prediction described in the embodiments.

[0660] The processor (160) can store the acquired bitstream in a storage (180).

[0661] The processor (160) can obtain prediction information from a bitstream. The processor (160) can obtain prediction information by performing entropy decoding on entropy-encoded prediction information of the bitstream.

[0662] At step (920), the processor (160) can derive a prediction mode for the target block.

[0663] The prediction mode may be one of the modes using prediction described in the embodiments.

[0664] For example, the prediction mode may be an intra prediction mode. The prediction for the target block may be an intra prediction mode, and the intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0665] The processor (160) can determine a prediction mode for the target block based on the prediction information.

[0666] For example, the prediction mode may be an intra prediction mode. The prediction for the target block may be an intra prediction mode, and the intra prediction mode of the intra prediction may be derived. One intra prediction mode may be determined from among the available intra prediction modes.

[0667] For example, the processor (160) can determine an intra prediction mode for a target block based on prediction information.

[0668] For example, the processor (160) can derive an intra prediction mode for the target block from among the available intra prediction modes for the target block.

[0669] At step (3630), the processor (160) can construct a reference sample for prediction.

[0670] The reference sample may be a reference sample used for prediction as described in the embodiments.

[0671] 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.

[0672] At step (940), the processor (160) can perform intra prediction on the target block using the derived intra prediction mode. The processor (160) can generate a prediction block for the target block through intra prediction.

[0673] In step (940), a prediction block can be generated by performing prediction on a target block using the derived prediction mode, and a reconstructed block, which is a sum of the prediction block and the reconstructed residual block, can be generated.

[0674] For example, in step (940), a prediction block can be generated by performing intra prediction on a target block using the derived intra prediction mode, and a restoration block, which is the sum of the prediction block and the restoration residual block, can be generated.

[0675] The embodiment may be combined with the operations of the decryption device (150) described above with reference to FIG. 1. For example, the operations of step (910) may be performed by the entropy decoder (161). The operations of steps (920), (930), and (940) may be performed by the predictor (163). Additionally, operations performed by other components of the decryption device (150) may be performed before, after, and between steps (910), (920), (930), and (940).

[0676]

[0677] The need for efficient prediction and encoding / decoding methods for efficient prediction

[0678] In embodiments, the image encoding / decoding method may use techniques such as histogram calculation and template matching for restored blocks / samples. However, these techniques may increase the complexity of encoding / decoding the image.

[0679] In embodiments, the video encoding / decoding method may construct a histogram for restored blocks / samples using a specific region. However, the limitation of this region may limit improvements in encoding / decoding efficiency.

[0680] In embodiments, a video encoding / decoding method may perform prediction on a luma / chroma block using a weighted sum of a limited number of prediction blocks, and may use a limited number of prediction candidates. However, the use of such a limited number of prediction candidates may limit improvements in encoding / decoding efficiency.

[0681] In embodiments, the video encoding / decoding method may construct a candidate list for a target block using the prediction modes of neighboring blocks. In other words, the candidate list for the target block may be constructed to include the prediction modes of the neighboring blocks.

[0682] For example, a video encoding / decoding method may construct a candidate list for a target block using the intra prediction modes of surrounding blocks. In other words, the candidate list for a target block may be constructed to include the intra prediction modes of surrounding blocks.

[0683] At this time, when an intra prediction technique other than a normal intra prediction mode (e.g., directional mode, planar mode, DC mode, etc.) is used for intra prediction of a surrounding block, when composing a candidate list of a target block, an alternative intra prediction mode corresponding to the other intra prediction technique used for the surrounding block may be used instead of the other intra prediction technique.

[0684] That is, the candidate list of the target block may include an alternative intra prediction mode corresponding to the above-mentioned other intra prediction technique instead of the other intra prediction technique of the surrounding block. 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 planar mode.

[0685] However, mapping to such fixed intra prediction modes may have limitations in improving encoding / decoding efficiency.

[0686] In embodiments, the video encoding / decoding method may use multiple signalings for template matching prediction. The use of these multiple signalings may limit improvements in encoding / decoding efficiency.

[0687] In embodiments, a video encoding / decoding method may perform prediction using fixed reference lines. In this case, multiple signaling methods may be used to select a reference line with a relatively large distance from the target block among the multiple available reference lines. The use of these multiple signaling methods may limit improvements in encoding / decoding efficiency.

[0688] In embodiments, the video encoding / decoding method may use samples within a luma block corresponding to a target block when performing prediction on a chroma block. The corresponding luma block may be a block having the same location as the target block. By using samples within the corresponding luma block, parallelization of prediction on the luma block and prediction on the chroma block may be limited.

[0689] In embodiments, the video encoding / decoding method may construct a histogram based on the amount of variation in pixel values ​​of restored pixels within a fixed region, and derive an intra prediction mode based on the histogram. The use of multiple such signaling methods may limit improvements in encoding / decoding efficiency.

[0690] The video encoding / decoding method described in the embodiments can overcome these limitations and have a configuration in which these limitations do not apply.

[0691] In embodiments, the image encoding / decoding method may extend the reference template construction step for analyzing reference samples. This extension may enable efficient encoding / decoding of images.

[0692] In embodiments, a video encoding / decoding method can perform prediction on a chroma component block using a plurality of pieces of information, and can improve prediction performance of the prediction.

[0693] In embodiments, the video encoding / decoding method may derive / sort intra prediction modes for analyzing reference samples, and may derive the intra prediction modes based on template matching costs. The complexity and signaling overhead of such intra prediction mode derivation / sorting methods may be reduced.

[0694] In embodiments, the image encoding / decoding method may extend the step of including surrounding blocks in the candidate list. This extension may enable efficient encoding / decoding of images.

[0695] In embodiments, the video encoding / decoding method may construct a candidate list of target blocks 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 planar mode, and a DC mode. Applying directionality to the non-directional mode may improve prediction performance.

[0696] In embodiments, a video encoding / decoding method may perform prediction on a chroma component block using block vector information. The use of block vectors may improve prediction performance.

[0697] In embodiments, the image encoding / decoding method may extend a reference line during the process of deriving an intra prediction mode through analysis of reference samples. Extending the reference line allows the image to be efficiently encoded / decoded.

[0698] By the methods described above, signaling overhead can be reduced and the time complexity required for prediction can be reduced.

[0699]

[0700] Prediction using guided prediction mode

[0701] Figure 10 is a flowchart of a prediction method according to an example.

[0702] The prediction method of the embodiments can be performed by the encoding device (110) and / or the decoding device (150).

[0703] For example, the encoding device (110) may perform the intra prediction method of the embodiment to compare the efficiencies of multiple prediction methods for a target block, and may perform the intra prediction method of the embodiment to generate a reconstructed block for the target block.

[0704] In one embodiment, the target block may be a CU, and may be one of the blocks described in the embodiments. For example, the target block may be at least one of a CTB, a CU, a PU, a TU, a sub-block, a block having a specified block size, and a block within a block size range defined in the present invention. Alternatively, the target block may represent a unit of coding. Alternatively, the target block may represent a specified region within the target picture.

[0705] Hereinafter, the size of the target block can be WxH. W can represent the width of the target block. H can represent the height of the target block. Each of W and H can be a positive integer. Each of W and H is 2 n can be. n can be a positive integer.

[0706] For example, the decryption device (150) may perform the intra prediction method of the embodiment to generate a restoration block for the target block.

[0707] Hereinafter, the processor may correspond to the processor (120) of the encoding device (110) and / or the processor (160) of the decoding device (150).

[0708] At step (1010), the processor can derive a prediction mode for the target block.

[0709] Step (1010) may correspond to step (810) described above with reference to FIG. 8 and step (920) described above with reference to FIG. 9. The content described in step (810) described above with reference to FIG. 8 or step (920) described above with reference to FIG. 9 may also be applied to step (1010).

[0710] For example, the prediction mode may be a mode that uses the prediction described in the embodiments.

[0711] For example, the prediction mode may be an intra prediction mode used in intra prediction.

[0712] The prediction mode for the target block can be derived using an artificial neural network.

[0713] For example, the intra prediction mode for a target block can be derived using an artificial neural network.

[0714] The processor can derive the intra prediction mode for the target block using one or more of the following methods: 1) a method using an intra prediction mode of a neighboring block (e.g., Most Probable Mode (MPM)); 2) a method deriving the intra prediction mode using reference sample prediction (e.g., Template-based Intra Mode Derivation (TIMD)); 3) a method deriving the intra prediction mode using reference template / sample analysis (e.g., Decoder side Intra Mode Derivation (DIMD)); 4) an intra template matching mode deriving the intra prediction mode based on a template (e.g., Intra Template Matching Prediction (IntraTMP)); and 5) a method deriving the intra prediction mode based on an intra prediction mode of a reference block (e.g., Occurrence-Based Intra Coding (OBIC)). Additionally, the methods related to prediction and / or prediction mode described in the embodiments can be used for deriving / encoding / decoding the intra prediction mode of the target block.

[0715] In embodiments, the intra template matching mode may derive the intra prediction mode based on the template matching cost.

[0716] In embodiments, a neighboring block may mean an adjacent block.

[0717] In embodiments, the intra prediction mode candidate list may be a Most Probable Mode (MPM) list. For example, the intra prediction mode for the target block may be derived using the MPM list.

[0718] An MPM list may contain one or more values. Each of the one or more values ​​may represent one or more intra prediction modes. Hereinafter, the terms "value" in the MPM list, "MPM" in the MPM list, "intra prediction mode" in the MPM list, and "candidate" in the MPM list may be used interchangeably.

[0719] 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 PDP) mode, and an LM mode (i.e., intra_fromLuma mode). In this case, the matrix-based intra prediction mode may mean a mode that defines matrix coefficients or weight values ​​based on a block size / shape and an intra prediction mode, and applies the matrix to a reference sample to make a prediction.

[0720] In intra prediction, when information about the intra prediction mode of a target block, such as an MPM list, is transmitted, intra prediction modes with the highest probability of matching the intra prediction mode of the target block can be selected as MPMs in the MPM list by referring to information about surrounding blocks of the target block. Through this selection, if the intra prediction mode used for intra prediction on the target block belongs to an MPM in the MPM list, a smaller number of binary bits can be used.

[0721] When a list of intra prediction modes for a target component of a target block is configured, the processor can use a method of deriving an intra prediction mode for the target component of the target block using an MPM, and can use a method of deriving an intra prediction mode for the target component of the target block using 1) a specific intra prediction mode and 2) an intra prediction mode for a reference component of the target block.

[0722] For example, the target component may be a chroma component. The reference component may be a luma component. Alternatively, the target component may be a luma component. The reference component may be a chroma component.

[0723] For example, the target component may be one of the three or four components described above, and the reference component may be another of the three or four components described above.

[0724] When the intra prediction mode of the target block is derived using the MPM list, the processor can perform reordering of 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.

[0725] Here, the context-adaptive MPM reordering method can predict 1) candidates of the MPM list and 2) reordering methods for the above candidates by using at least one of information of pre-restored surrounding blocks and information of the target block.

[0726] Through such prediction, the intra prediction mode for the target block can be directly derived using at least one of information of the target block and information of the pre-restored surrounding blocks, without signaling information about the prediction mode for the target block.

[0727] At step (1020), the processor can construct a reference sample for prediction for the target block.

[0728] Step (1020) may correspond to step (820) described above with reference to FIG. 8 and step (930) described above with reference to FIG. 9. The content described in step (820) described above with reference to FIG. 8 or step (930) described above with reference to FIG. 9 may also be applied to step (1020).

[0729] The processor can construct the 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 the reference sample for sample analysis.

[0730] Additionally, methods related to reference samples in the embodiments may be used to construct 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.

[0731] At step (1030), the processor can perform prediction on the target block using the derived prediction mode.

[0732] Step (1030) may correspond to step (830) described above with reference to FIG. 8 and step (940) described above with reference to FIG. 9. The content described in step (830) described above with reference to FIG. 8 or step (940) described above with reference to FIG. 9 may also be applied to step (1030).

[0733] The processor can perform prediction on a target block using the derived prediction mode. Alternatively, the processor can perform intra prediction on a target component of the target block using the derived intra prediction mode.

[0734] The processor can generate a prediction block for the target block by making a prediction for the target block.

[0735] In one embodiment, the processor is configured to perform one or more of: 1) non-directional intra prediction; 2) directional intra prediction; 3) inter-color prediction; 4) sub-block-by-block prediction; 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 multiple reference lines based on templates (e.g., Template-based Multiple Reference Line (TMRL)); and 9) prediction using an intra-prediction mode based on an extrapolation filter (e.g., Extrapolation filter-based intra prediction (EIP) mode). 10) Intra prediction for a target block can be performed using one or more methods of matrix-based intra prediction.

[0736] Additionally, the processor can perform corrections on the predicted samples. The predictions may be intra-predictions. Methods related to filtering / sampling / interpolation for the samples described in the embodiments can be applied to the predicted samples.

[0737] The processor can perform filtering on prediction samples of a prediction block during the process of performing intra prediction.

[0738]

[0739] Decoder-side Intra Mode Derivation (DIMD)

[0740] When DIMD is applied, up to five intra-modes can be derived from the reconstructed neighboring samples, and these five predictors can be combined with a non-directed mode predictor (a planar or block vector predictor) with weights derived from a histogram of gradients. The selection between the non-directed modes can be made based on the template cost. Specifically, the block vectors of all adjacent and non-adjacent merge candidates (coded as IntraTMP or IBC) can be compared with the planar predictions on the reconstructed template. The template cost (SAD, MR-SAD, or SATD) can be used to select the best predictor among the predictors.

[0741] Division operations in weight derivation can be performed using the same lookup table (LUT)-based integerization scheme used by CCLM. For example, the division operation in the orientation calculation in [Equation 1] below can be computed using the LUT-based schemes in [Equations 2] through [Equations 5]:

[0742] [Formula 1]

[0743] Orient = G y / G x

[0744] [Formula 2]

[0745] x = Floor( Log2( Gx ) )

[0746] [Formula 3]

[0747] normDiff = ( ( Gx << 4 ) >> x ) & 15

[0748] [Formula 4]

[0749] x += ( 3 + ( normDiff != 0 ) ? 1 : 0 )

[0750] [Formula 5]

[0751] Orient = (Gy * ( DivSigTable[ normDiff ] | 8 ) + ( 1<<( x-1 ) )) >> x

[0752] Here, DivSigTable can be as shown in [Formula 6] below.

[0753] [Formula 6]

[0754] DivSigTable

[0016] = { 0, 7, 6, 5 ,5, 4, 4, 3, 3, 2, 2, 1, 1, 1, 1, 0}.

[0755] For a block of size W × H, the weights of each of the five induced modes can be modified if either the upper histogram magnitudes or the left histogram magnitudes are more than twice the other. In this case, the weights are location-dependent and can be computed as follows:

[0756] If the upper histogram is twice as large as the left histogram, [Equation 7] below can be used.

[0757] [Formula 7]

[0758]

[0759] If the left histogram is twice as large as the upper histogram, [Equation 8] below can be used.

[0760] [Formula 8]

[0761]

[0762] Here, wDIMD i can be the unmodified uniform weight of the selected DIMD. Δ i can be pre-defined and set to 10.

[0763] The derived intra modes can be included in the list of intra MPMs. Therefore, the DIMD process can be performed before the MPM list is built. The derived intra modes of a DIMD block can be stored with the block and used to build the MPM lists of neighboring blocks.

[0764] Finally, the regions of neighboring reconstructed samples used to compute the histogram of gradients can be modified depending on the availability of reconstructed samples. The region of decoded reference samples of the current W × H luma CB can be extended by up to W additional columns toward the top-right side, if available. The region can be extended by up to H additional rows toward the bottom-left side, if available.

[0765]

[0766] DIMD chroma mode

[0767] Figure 11 shows neighboring reconstructed samples used for DIMD chroma mode according to an example.

[0768] The DIMD chroma mode can derive a chroma intra prediction mode of a current block based on neighboring reconstructed Y, Cb, and Cr samples in the second neighboring row and column, as shown in FIG. 11. Specifically, for each collocated reconstructed luma sample of the current chroma block, a horizontal gradient and a vertical gradient can be computed together with the reconstructed Cb and Cr samples to build a histogram of gradients (HoG). Next, the intra prediction mode with the largest histogram amplitude values ​​can be used to perform chroma intra prediction of the current chroma block.

[0769] When the intra prediction mode derived from the DIMD chroma mode is the same as the intra prediction mode derived from the DM mode, the intra prediction mode with the second largest histogram amplitude value can be used as the DIMD chroma mode. A CU level flag can be signaled to indicate whether the DIMD chroma mode is applied.

[0770] The luma region of the reconstructed samples for computing the histogram of gradients for the chroma DIMD mode can be determined within the luma block. For example, the histogram can be derived by computing the gradients from the reconstructed samples corresponding to the second and third lines within the corresponding luma block.

[0771]

[0772] Template-based Intra Mode Derivation (TIMD)

[0773] TIMD may be a method of performing intra prediction based on at least one intra prediction mode for a template composed of reconstructed samples adjacent to a current block, and deriving one or more intra prediction modes.

[0774] At this time, a candidate list containing one or more intra prediction modes for the current block can be derived.

[0775] The above template may be N adjacent reference sample lines, which may vary depending on the size of the current block. N may be at least one of 1, 2, 3, or 4.

[0776] A prediction block can be generated by performing intra prediction based on the candidate intra modes in the above list for the above template. Thereafter, by comparing the errors between the predicted samples and the restored samples, at least one intra prediction mode with a relatively smaller error can be derived.

[0777] The above error can be derived based on at least one of the methods for deriving statistical values ​​described in the embodiments, such as SAD, MR-SAD, SATD, etc.

[0778] Combined (blended or weighted sum) prediction can be performed using the above-described N modes. The weights for combined prediction can be derived based on the above-described error values.

[0779] The derived intra modes can be included in the list of intra MPMs. Therefore, the TIMD process can be performed before the MPM list is built. The derived intra modes of a TIMD block can be stored with the block and used to build the MPM lists of neighboring blocks.

[0780]

[0781] Intra template matching

[0782] Figure 12 shows a search area for intra template matching according to an example.

[0783] Intra Template Matching Prediction (IntraTMP) may be an intra prediction mode that copies a reconstructed block with a template most similar to the current template from a reconstructed portion of the current frame and uses it as a prediction block. For a predefined search range, the encoder can search for the template most similar to the current template within the reconstructed portion of the current frame and use the corresponding block as a prediction block. The encoder can then signal the use of this mode, and the same prediction operation can be performed at the decoder side. At this time, information indicating the location from the current block to the predicted block can be referred to as a block vector (BV) and can be stored for the current block.

[0784] A prediction signal can be generated by matching an L-shaped, top-only, or left-only template of the current block with another block within a predefined search area of ​​FIG. 12. There can be six predefined search areas, R1 to R6 in FIG. 12. R1 to R6 can include portions of reconstructed samples within the current CTU located above, left, below-left, and above-right of the current block, along with reconstructed samples from the upper and left CTUs.

[0785] IntraTMP can apply implicit merge mode and derive merge lists. In implicit merge mode, merge candidates can be considered without signaling a merge flag or index. In particular, reference locations pointed to by the block vectors of all adjacent and non-adjacent merge candidates (coded in IntraTMP or IBC mode) can be used as additional candidates beyond the default search areas. The same template matching cost can be used to compare merge locations and default locations. For bidirectional IBC merge candidates, two candidates corresponding to each reference frame can be retained. Similarly, for IntraTMP, two candidates can be considered: the best candidate by template search and the coded candidate.

[0786] Sum of Absolute Differences (SAD), MR-SAD or SATD can be used as a cost function for template matching cost.

[0787] A given search order of six regions, i.e., R4, R5, R6, R1, R3, and R2, can be utilized. Within each region, the decoder can build a candidate list of 19 template-matching block vectors by sorting them in ascending order of template cost. The following modes can be supported:

[0788] 1. Single predictor: A single predictor can be selected from the candidate list.

[0789] 2. Blending of multiple predictors: Multiple predictors can be blended to derive a final prediction block. Blending weights can be calculated from the template matching cost of each predictor or using a Wiener filter-based weight derivation method.

[0790] 3. Sub-pel precision: When a single predictor is used, sub-pel precision can be used with 1 / 2-pel precision, 1 / 4-pel precision, and 3 / 4-pel precision, each with 8 possible directions.

[0791] 4. Linear filter model: A linear filter model can be derived between the reference template and the current template and applied to the linear model as a reference block. This mode can be used for a single predictor when sub-pel precision is not used.

[0792] Additionally, IntraTMP with Local Illumination Compensation (LIC) may be permitted. The following considerations may be taken:

[0793] 1. The use of LIC and filter-based linear models (FLM) (CCCM-like filtering) may be mutually exclusive for a given CU.

[0794] 2. The use of LICs with fusion within IntraTMP may be permitted.

[0795] 3. The use of top-only and left-only templates for LIC model determination may be permitted for screen content coding. For camera-capture coding, only the top-left template may be employed.

[0796] 4. For screen content coding, the Multi Mode Linear Model (MMLM) can be supported similarly to IBC-LIC.

[0797] When LIC is used for a given CU, the Intra TMP search process can utilize Mean Removal SAD (MR-SAD) rather than the SAD distortion function.

[0798] The search regions (SearchRange_w, SearchRange_h) can be set in proportion to the block sizes (BlkW, BlkH) to compare a fixed number of SADs per pixel. That is, the setting can be performed according to [Equation 9] and [Equation 10] below.

[0799] [Formula 9]

[0800] SearchRange_w = min(64,a * BlkW)

[0801] [Formula 10]

[0802] SearchRange_h = min(64,a * BlkH)

[0803] Here, 'a' can be a constant that controls the gain / complexity trade-off. In practice, 'a' can be equal to 5.

[0804] To speed up the template matching process, template matching can be performed at four-sample intervals across the search area. Furthermore, after deriving the optimally matching sample, a refinement process can be performed. That is, refinement can be performed through a secondary template matching search around the optimally matching sample.

[0805] Intra template matching can be enabled for CUs having a size of 64 or less in width and height. The maximum CU size for such intra template matching can be configurable.

[0806] A flag corresponding to whether intra template matching prediction mode is used can be signaled at the CU level.

[0807]

[0808] Block vector candidates derived from intra-TMP for IBC

[0809] Figure 13 illustrates the use of IntraTMP block vectors for IBC blocks.

[0810] Block vectors (BVs) derived from intra-template matching prediction (IntraTMP) can be used for intra-block copying (IBC). The IntraTMP BVs of stored neighboring blocks can be used as spatial BV candidates in constructing an IBC candidate list, along with the IBC BVs.

[0811] IntraTMP block vectors can be stored in an IBC block vector buffer, and the current IBC block can use the IBC BVs and IntraTMP BVs of neighboring blocks as BV candidates in the IBC BV candidate list of the current block, as shown in Fig. 13. IntraTMP block vectors can be stored in quarter-pel resolution for coding IBC block vectors and HMVP.

[0812]

[0813] Intra Prediction Fusion

[0814] This intra prediction method can derive a final prediction block by applying weights to multiple prediction blocks generated from different reference lines and synthesizing (combining) them. The process of deriving the prediction blocks to be used in the synthesis process can be described as follows:

[0815] 1) For directional (or angular) intra prediction modes including single mode cases of TIMD and DIMD, the proposed method can derive intra prediction by applying weights to intra predictions obtained from multiple reference sample lines as expressed in [Equation 10] below.

[0816] [Formula 10]

[0817] P fusion = w0P line + w1P line+1

[0818] Here, P line can be an intra prediction using the default reference line, and P line+1 can be a prediction using the line above the default reference line. The weights can be set as shown in [Equation 11] below.

[0819] [Formula 11]

[0820] w0= 3 / 4, w1= 1 / 4

[0821] 2) For the TIMD mode that performs joint prediction, P line can be used for the first TIMD mode (w0= 1, w1= 0), and P line+1 can be used for the second TIMD mode (w0= 0, w1= 1). In this case, both the first and second TIMD modes are directional prediction modes, and the intra prediction synthesis method can be performed when the difference between the angles or mode values ​​of the two modes is greater than a predetermined threshold. Alternatively, whether to apply the prediction synthesis method can be determined based on the degree of separation from the horizontal / vertical mode.

[0822] 3) For the DIMD mode that performs joint prediction, the number of prediction blocks selected for weighted joint prediction may be N (e.g., 6).

[0823] The intra prediction synthesis method can be performed in the case of intra directional prediction that applies an interpolation filter, or can be applied to a luma block when the current block size is larger than a predetermined size (e.g., 16).

[0824]

[0825] Adaptive Histogram of Gradient (HoG) for Decoder-side Intra Mode Derivation (DIMD)

[0826] In one embodiment, a method for calculating HoG in DIMD may be used.

[0827] Instead of using a fixed DIMD template, an adaptive number of samples can be used to compute the HoG based on histogram amplitudes.

[0828] A DIMD template consisting of restored samples further away from the current block can be used.

[0829] Additionally, the amplitude can be scaled by a factor that depends on the distance from the current block and the computed directions.

[0830] The number of DIMD predictors can be scaled for larger blocks.

[0831] When DIMD is used, HoG can be calculated by analyzing samples of the DIMD template.

[0832] A fixed template consisting of three lines at the top and left of the current block can be used.

[0833] When DIMD is used, N (e.g., 5 or 7) DIMD modes can be derived by analyzing the directionality of samples within the template region.

[0834] In deriving HoG, a Sobel filter can be applied to samples within the middle line of the template region to derive horizontal / vertical gradients, and based on these, directionality and directional prediction modes can be derived. Furthermore, the amplitude can be derived as the sum of the absolute values ​​of the derived horizontal / vertical gradients, and the HoG can be derived by accumulating the amplitude for each directional prediction mode.

[0835] Based on the derived HoG, one or more intra prediction modes can be selected.

[0836] Then, the predictors (prediction modes or prediction blocks) obtained with the DIMD modes can be blended to form the final DIMD predictor. Uniform blending or spatial blending can be used, where the DIMD predictors derive weights based on the relative amplitude values ​​of the modes within the HoG, and can be combined with a planar predictor or a block-vector-based predictor.

[0837] In an embodiment, instead of using a fixed template, the number of samples to which the Sobel filter is applied can be adaptively determined. That is, during the process of deriving amplitudes for samples within the template and accumulating the amplitudes for each directional prediction mode, the filter application and HoG derivation process can be stopped when the accumulated value reaches a given threshold.

[0838] Meanwhile, if the above threshold is not satisfied, the DIMD template can be extended to include more lines on the top and left of the current block.

[0839] This process can be performed independently for templates above, to the left, and above-left of the current block.

[0840] The threshold used to fill the HoG can be determined depending on the block size.

[0841] When a given gradient is computed for a reference sample, the amplitude of the reference sample may be scaled with weights that depend on the distance from the current block to the reference sample and the direction of the neighboring template sample.

[0842] Additionally, the DIMD process can select more intra modes from the HoG when the current block size is a certain size (e.g., when the width × height size is 128). For example, if the current block has an area larger than 128, up to 8 DIMD predictors can be blended. On the other hand, for smaller blocks, the number of predictors can be maintained at 6, as in the existing method.

[0843] In embodiments, the number of DIMD predictors may be maintained at 6 for all block sizes. No scaling may be used for amplitude in the HoG calculation.

[0844]

[0845] Matrix-based intra prediction to replace existing conventional intra prediction modes.

[0846] Figure 14 shows L-shaped surrounding samples for a prediction block according to an example.

[0847] A matrix of weights defined for block shape and intra prediction mode can be used.

[0848] These weights can be multiplied by surrounding reference templates to derive prediction samples that replace the usual intra prediction.

[0849] Weights can be applied to reference samples of causal neighbor templates in the form of L, as illustrated in Fig. 14.

[0850] Reference samples within a causal neighborhood can be denoted by r. F(x, y) can be a matrix of weights.

[0851] The predicted sample P(x, y) can be derived as shown in [Equation 12] below.

[0852] [Formula 12]

[0853]

[0854] Here, k can represent the index of the reference sample within the template.

[0855] The prediction of the embodiment can be used for block sizes with width and height up to 32. Here, the prediction may not be used for block sizes 4x32, 32x4, 8x32 and 32x8.

[0856] For blocks with width and height up to 16, the template size can be 2. In this case, prediction can only be used for intra prediction modes 0, 1, and (2 + 2*k).

[0857] k can be an integer greater than or equal to 0, meaning that predictions can be used for planar mode, DC mode, and even-numbered intra prediction modes.

[0858] For other blocks, the template size can be set to 1. In this case, prediction can only be used for intra prediction modes 0, 1, and (2 + 4*k).

[0859] Prediction can only be performed on 16x16 locations. The remaining samples can be generated by bilinear interpolation.

[0860] For all block sizes, symmetry based on block shape and mode can be used.

[0861] For intra prediction modes greater than 18 and less than 50, the reference length can be set to W and H. For other intra prediction modes, the reference length can be set to 2*W and 2*H.

[0862]

[0863] Configuration of templates for reference template / sample analysis

[0864] An intra prediction mode can be derived using analysis of a reference template / sample. The derived intra prediction mode can be used as a candidate in a candidate list and as the intra prediction mode of a target block.

[0865] The template may be a restored area around the target block.

[0866] For example, analysis of the reference sample may include decoder-side intra-mode derivation (DIMD).

[0867] Analysis of reference samples of the target block can be applied in the process of deriving the intra prediction mode.

[0868] Analysis of the reference samples of the target block may be an analysis of the directionality of the pre-encoded / decoded reference samples with respect to the template of the target block.

[0869] This analysis can extract orientation and magnitude information by analyzing vertical and horizontal gradients between a target sample and surrounding samples. The extracted information can be used to generate a histogram of gradients (HoG).

[0870] In embodiments, HoG may mean a histogram of oriented gradient.

[0871] In embodiments, magnitude may mean size or amplitude. In embodiments, the terms "magnitude," "size," and "amplitude" may be interchangeable.

[0872] For example, in a HoG, bins can be generated, each corresponding to an available intra prediction mode. A magnitude of each bin can be generated. The magnitude of a bin can represent the magnitude of the intra prediction mode corresponding to the bin. The available intra prediction modes can include directional intra prediction modes.

[0873] Samples for these analyses can be constructed using the methods described below.

[0874] The reference sample line described with reference to FIG. 3 may be a line of restored samples. In embodiments, "reference line" and "restored sample line" may refer to a "reference sample line." In other words, the description of a "reference sample line" may also apply to a "reference line" and a "restored sample line."

[0875] The indexes for the aforementioned reference sample lines and the indexes for the restoration sample lines may be used differently. For example, the index of the restoration sample lines may start from 1. Restoration sample line 1 may refer to a restoration sample line adjacent to the target block. The restoration sample line may correspond to reference sample line 0. Additionally, restoration sample line n may correspond to reference sample line n-1.

[0876] Samples for analysis may be constructed using one or more reconstructed samples or a combination of samples surrounding the target block. Filtering may be applied to the constructed samples.

[0877] Here, each of the restored samples on multiple restored sample lines can be used as is. Alternatively, filtering can be performed on samples on the same restored sample line or between samples on different restored sample lines, and samples for analysis can be constructed after such filtering.

[0878] In embodiments, ref[m, n] may represent a sample configured for analysis.

[0879] In embodiments, rec[m, n] may be a restored sample around the target block or a sample generated by applying filtering to the restored sample above.

[0880] In embodiments, a sample generated by applying filtering to a specific sample may be referred to as a filtered specific sample. In other words, rec[m, n] may be a restored sample around the target block; or a filtered restored sample.

[0881] Each of m and n can be an integer and can represent the coordinates of the sample.

[0882] If the horizontal and vertical sizes of the target block are W and H, respectively, the position of the leftmost-topmost sample within the target block may be (0, 0). Based on the position of this leftmost-topmost sample, the relative position of the upper left reference sample closest to the leftmost-topmost sample may be (-1, -1). In other words, the position of the reference sample diagonally adjacent to the leftmost-topmost sample of the target block may be (-1, -1).

[0883] Samples for analysis can be constructed using restoration sample lines around the target block.

[0884] At this time, samples within the sample line adjacent to the target block can be used.

[0885] Alternatively, samples within a sample line that is not adjacent to the target block may be used.

[0886] Alternatively, adjacent and non-adjacent samples to the target block may be used together.

[0887] Alternatively, samples indicated by the information described in the examples may be used.

[0888] After a reference template for analysis is configured, all samples within the configured reference template can be used for analysis.

[0889] Alternatively, after a reference template for analysis is constructed, some samples within the constructed reference template can be used for analysis.

[0890] Average fusion can be performed to construct a sample for analysis. Average fusion can be a method of averaging two or more different reconstructed sample lines in units of samples. By average fusion, a (fused) reconstructed sample line that is the average of the units of samples of two or more different reconstructed sample lines can be derived, and the derived (fused) reconstructed sample line can be used as a reconstructed sample line for the target block described in the embodiments.

[0891] In one embodiment, average fusion may be performed on two restored sample lines.

[0892] For example, average fusion between restored sample line 1 and restored sample line 1+k can be performed, where k can be a positive integer greater than or equal to 1. The maximum value of K can be M.

[0893] In embodiments, M may be the number of restored sample lines.

[0894] For example, average fusion of restored sample lines 1+p and restored sample lines 1+q can be performed. Here, each of p and q can be a positive integer greater than or equal to 1. The maximum value of each of p and q can be M. p and q can be different from each other.

[0895] Average fusion can be performed on three restored sample lines.

[0896] For example, average fusion of restored sample line 1, restored sample line 1+k, and restored sample line 1+l can be performed. Here, k can be a positive integer greater than 1. l can be a positive integer greater than k. The maximum values ​​of k and l can be M.

[0897] For example, average fusion of the restored sample line Ak, the restored sample line A, and the restored sample line A+l can be performed. Here, each of k and l can be a positive integer greater than or equal to 1. Furthermore, Ak can be a positive integer greater than or equal to 1. A+l can be a positive integer less than or equal to M.

[0898] Weighted sum fusion may be performed to construct a sample for analysis. Weighted sum fusion may be a method of combining two or more different reconstructed sample lines by applying weights in units of samples. By weighted sum fusion, a (fused) reconstructed sample line that is a weighted sum in units of samples of two or more different reconstructed sample lines may be derived, and the derived (fused) reconstructed sample line may be used as a reconstructed sample line for the target block described in the embodiments.

[0899] As described above, reconstructed samples can be constructed by average fusion, and the samples constructed by average fusion can be used as samples for analysis. Here, the reconstructed samples constructed by average fusion can be identical to the reconstructed samples constructed by weighted sum fusion. The reconstructed samples constructed by weighted sum fusion can be used as samples for analysis. The description of average fusion in the embodiments can also be applied to weighted sum fusion. The description of weighted sum fusion in the embodiments can also be applied to average fusion.

[0900] In one embodiment, weighted sum fusion may be performed on two restored sample lines. Weighted sum fusion may be performed as shown in [Equation 13] below.

[0901] [Formula 13]

[0902] Predicted fusion sample = Weight 1 × Sample of restored sample line A + Weight 2 × Sample of restored sample line B

[0903] The predicted fusion sample may be a reference sample generated by fusion.

[0904] For example, {weight1, weight2} can be one of {0.6, 0.4}, {0.7, 0.3}, {0.8, 0.2}, and {0.9, 0.1}.

[0905] For example, weighted sum fusion can be performed on a line of three restored samples. Weighted sum fusion can be performed as shown in [Equation 14] below.

[0906] [Formula 14]

[0907] Predicted fusion sample = Weight 1 × Sample from restored sample line A + Weight 2 × Sample from restored sample line B + Weight 3 × Sample from restored sample line C

[0908] For example, {weight1, weight2, weight3} can be one of {0.4, 0.3, 0.3}, {0.5, 0.25, 0.25}, {0.6, 0.2,0.2}, {0.7, 0.15, 0.15}, and {0.8, 0.1, 0.1}.

[0909] Weights can be defined using coding parameters and / or properties of the object. For example, weights can be determined using one or more of slice type (e.g., which slice among I / P / B slices is used for the target block), signal component (e.g., which component of the target block is Y / U / V), quantization parameter (QP), block size, and block shape.

[0910] The method of configuring the upper reference sample, which constitutes the upper reference sample located at the top of the target block, and the method of configuring the left reference sample, which constitutes the left reference sample located at the left of the target block, may be different from each other.

[0911] Here, the upper reference sample and the left reference sample can be constructed from two different restored sample lines among the multiple restored sample lines.

[0912] For example, the left reference sample may be constructed using restored sample line A, and the top reference sample may be constructed using restored sample line B.

[0913] Here, A and B can be different integers greater than or equal to 1.

[0914] Here, the left reference sample and the top reference sample can be constructed using one selected from average fusion and weighted sum fusion.

[0915] For example, reference samples generated by average fusion and reference samples generated by weighted sum fusion can be used for reference samples at each location.

[0916] For example, for each location, a reference sample generated by average fusion and a reference sample without fusion applied can be used.

[0917] For example, reference samples generated by weighted sum fusion and reference samples to which fusion is not applied can be used for reference samples at each location.

[0918] A reference sample may be constructed using one or more of the methods of constructing a reference sample of the embodiments.

[0919] Here, one or more restored sample lines, excluding a restored sample line having a value identical to or similar to a value of an adjacent restored sample line among a plurality of restored sample lines, can be used to construct a reference sample.

[0920] In embodiments, the adjacent restoration sample line may be a restoration sample line adjacent to the target block. For example, the adjacent restoration sample line may be the reference sample line 0 described above with reference to FIG. 3.

[0921] That is, a similar restored sample line may be excluded from the restored sample lines used to construct a reference sample. A similar restored sample line may be a restored sample line having a value that is identical to or similar to the value of an adjacent restored sample line.

[0922] In one embodiment, the values ​​of the restored sample lines may be sum of absolute differences (SAD) values.

[0923] For example, a similar restored sample line may have a SAD value that is equal to or similar to the SAD value of an adjacent restored sample line.

[0924] After a restoration sample line having a SAD value that is the same as the SAD value of an adjacent restoration sample line among a plurality of restoration sample lines or having a SAD value whose difference from the SAD value of the adjacent restoration sample line is less than a certain value is excluded, a reference sample can be constructed using the remaining restoration sample lines.

[0925] Here, SAD values ​​between reference sample candidates and restored sample line 1 can be calculated. The reference sample can be reconstructed by preferentially placing candidates having a large SAD value with restored sample line 1 among the reference sample candidates.

[0926] In embodiments, the reference sample candidates may be one or more reference samples described as being used for composing the reference sample. For example, the reference sample candidates may be reference samples within one or more restored sample lines described as being used for composing the reference sample.

[0927] Reference sample candidates may include multiple reference sample lines. A reference sample candidate may mean a reference sample line.

[0928] SAD values ​​between multiple reference sample candidates and restored sample line 1 can be calculated. The SAD value of the reference sample candidate can mean the SAD value between the reference sample candidate and restored sample line 1.

[0929] Reference sample candidates can be reconstructed by preferentially placing reference sample candidates with large SAD values ​​among multiple reference sample candidates. Reconstructing the reference sample candidates may mean excluding some of the multiple reference sample candidates; and changing the order of the multiple reference sample candidates.

[0930] When reference sample candidates are reconstructed, reference sample candidates having a SAD value equal to the SAD value of other reference sample candidates and reference sample candidates having a SAD value whose difference from the SAD value of other reference sample candidates is less than or equal to a reference value may be excluded.

[0931] Here, the SAD value between the restoration sample candidates and the restoration sample line 1 can be calculated. After the restoration sample lines having the same SAD value as the SAD value of the adjacent restoration sample line or having a SAD value whose difference from the SAD value of the adjacent restoration sample line is less than a certain value are excluded, the reference sample can be reconstructed using the remaining restoration sample lines.

[0932] In addition to SAD, other statistical values ​​described in the examples may be used as criteria for obtaining values ​​of the restored sample line and reference sample candidates.

[0933] In the embodiments, the value / number / index of the restored sample line used for analysis may be a value preset in the encoding device (110) / decoding device (150), and may be a value signaled from the encoding device (110) to the decoding device (150).

[0934] It can be determined whether information of reference samples for intra prediction is signaled using the reconstructed blocks surrounding the target block.

[0935] For example, an index for a restoration sample line that can be used can be assigned using restoration blocks surrounding the target block. In this case, information indicating which sample is used for analysis may not be signaled.

[0936] For example, a flag could be signaled as to whether adjacent restoration sample lines are used.

[0937] The coding parameters described in the examples can be used to derive reference samples.

[0938] The composition of the sample for analysis against the reference template can be performed using coding parameters and / or properties of the object. For example, the composition of the sample can be determined using one or more of the slice type (e.g., which slice among I / P / B slices is used for the target block), signal component (e.g., which component of the target block is Y / U / V), QP, block size, and block shape.

[0939]

[0940] Analysis of encoded / decoded blocks

[0941] In embodiments, the target block may be a block that is the subject of analysis and may be an encoded / decoded block having restoration samples.

[0942] For example, the target block may be the current block that is the target of encoding / decoding. In this case, the analysis of the embodiment may be performed after encoding / decoding of the target block is performed.

[0943] For example, the target block may be a restoration block or a reference block used for encoding / decoding the current block. In this case, the analysis of the embodiment may be performed when encoding / decoding the current block is performed.

[0944]

[0945] Figure 15 shows the area of ​​samples used for analysis within a target block according to an example.

[0946] Figure 16 shows the area of ​​samples used for analysis determined according to the size of the target block according to an example.

[0947] In one embodiment, an intra prediction mode may be determined using analysis of samples of a region within a decoded target block; and a region including the target block and surrounding regions. This intra prediction mode may be used for prediction for the target block.

[0948] Additionally, the target block may be encoded / decoded using at least one of the prediction modes described in the embodiments, such as intra prediction, inter prediction, template matching prediction, and intra block copy (IBC). In such encoding / decoding, an intra prediction mode determined using analysis of a decoded sample described in the embodiments may be used.

[0949] This analysis can extract orientation and magnitude (or size) information by analyzing vertical and horizontal gradients between a target sample and surrounding samples. The extracted information can be used to generate a histogram of gradients (HoG). Additionally, other geometric information between the target sample and surrounding samples can be extracted, and the extracted geometric information can be used to generate a HoG.

[0950] In embodiments, HoG may mean a histogram of oriented gradient.

[0951] In one embodiment, after decoding of a target block is completed, N or more intra prediction modes can be derived by analyzing samples within the target block, as illustrated in FIG. 15. N can be an integer greater than or equal to 1.

[0952] For example, intra prediction modes can be derived using HoG.

[0953] For example, the intra prediction mode may be the intra prediction mode with the largest accumulated magnitude value in the HoG.

[0954] N intra prediction modes may be N intra prediction modes having relatively large cumulative magnitudes on the HoG.

[0955] Here, N can be an integer greater than 0. For example, N can be 5.

[0956] The intra prediction mode can be selected based on the information of each directional intra prediction mode based on the HoG.

[0957] After decoding of the target block is completed, N or more intra prediction modes can be derived by analyzing a region composed of samples within the target block and surrounding blocks, as illustrated in FIG. 16. N can be an integer greater than or equal to 1.

[0958] For example, intra prediction modes can be derived using HoG.

[0959] For example, the intra prediction mode may be the intra prediction mode with the largest accumulated magnitude value in the HoG.

[0960] N intra prediction modes may be N intra prediction modes having relatively large cumulative magnitudes on the HoG.

[0961] Here, N can be an integer greater than 0. For example, N can be 5.

[0962] The intra prediction mode can be selected based on the information of each directional intra prediction mode based on the HoG.

[0963]

[0964] Composition of the sample for analysis

[0965] The sample used for analysis can be configured according to the embodiments described below.

[0966] In one embodiment, analysis can be performed using samples within a specific region, such as that illustrated in FIG. 15.

[0967] In Fig. 15, W may represent the horizontal size of the target block. H may represent the vertical size of the target block.

[0968] Among the entire regions of the target block, the samples used for analysis may be limited to samples within a specific region of size M × N.

[0969] M can be an integer greater than or equal to 1 and less than or equal to W. N can be an integer greater than or equal to 1 and less than or equal to H.

[0970] The specific region may be the lower-right region within the target block. The rightmost and bottommost sample of the target block may be the rightmost and bottommost sample within the specific region. In other words, the specific region may contain the target block, and the target block may be located at the lower-right corner of the specific region.

[0971] Each of M and N can be derived using the coding parameters described in the embodiments.

[0972] For analysis, all samples from a specific area within the target block are available.

[0973] For analysis, K samples from a specific area within the target block can be used. K can be an integer greater than or equal to 1.

[0974] In one embodiment, analysis can be performed using samples within a specific region, such as that illustrated in FIG. 16.

[0975] In Fig. 16, W may represent the horizontal size of the target block. H may represent the vertical size of the target block. M may be a value for determining the horizontal size of a specific area used for analysis. N may be a value for determining the horizontal size of a specific area used for analysis.

[0976] M can be a value greater than or equal to 1 and less than or equal to W. N can be a value greater than or equal to 1 and less than or equal to H.

[0977] The samples used for analysis can be restricted to those within a specific region determined using M and N.

[0978] For example, a particular area may be an area of ​​size M × N.

[0979] For example, a particular region may be an area of ​​size (2W - M) × (2H - N).

[0980] For example, a specific region may contain a target block. The rightmost and bottommost samples of the target block may be the rightmost and bottommost samples within the specific region. In other words, the specific region may be part of the target block and may be located at the right-bottom of the target block.

[0981] Each of M and N can be derived using the coding parameters described in the embodiments.

[0982] For analysis, all samples in a specific area within the target block location can be used.

[0983] For analysis, K samples from a specific area within the target block location can be used. K can be an integer greater than or equal to 1.

[0984] Once samples for analysis are constructed by the method described above, filtering can be applied to the samples.

[0985] For example, the restored samples can be used as is (without filtering).

[0986] For example, samples can be reconstructed by filtering between samples on the same restored sample line.

[0987] For example, samples can be reconstructed by filtering between samples on different restored sample lines.

[0988] For example, the filtering process described in the embodiments can be applied to the sample.

[0989]

[0990] Use of directional modes derived from analysis

[0991] The modes derived using the method of the embodiments can be stored within the block in which encoding / decoding is completed.

[0992] The derived modes can be N modes. The derived modes can be directional modes. That is, N directional modes can be specified for a block for which encoding / decoding is completed. Alternatively, N directional modes can be determined for a block for which encoding / decoding is completed.

[0993] The encoded / decoded block in which these N directional modes are stored may be a block described in the embodiments, such as a target block, a reference block, and a surrounding block.

[0994] N can be an integer greater than or equal to 1.

[0995] The N directional modes derived using the method of the embodiments can be used in the processes described below.

[0996] In one embodiment, when a block is encoded / decoded using a specific intra prediction technique and / or a specific inter prediction technique, information about the derived modes may be used. For example, the derived modes may be used as intra prediction modes provided from the block.

[0997] For example, the derived mode may replace a specific intra-prediction mode used in a specific intra-prediction technique and / or a specific inter-prediction technique. The derived mode may replace a specific intra-prediction mode derived from a specific intra-prediction technique and / or a specific inter-prediction technique.

[0998] For example, a particular intra prediction technique may include the construction of a (candidate) list, the construction of an MPM list, the construction of a TIMD candidate list, the construction of a TMRL intra prediction mode candidate list, and the construction of an intra prediction mode (IPM) list, as described in the embodiments.

[0999] For example, the derived modes can be used as candidates for the list described in the embodiments. Combinations of the derived modes can be used as candidates for the list described in the embodiments.

[1000] For example, the derived mode can be used as a replacement for the above candidate when the candidate from the list described in the embodiments is used.

[1001] For example, the derived mode may be used in place of a candidate from the list described in the embodiments when the candidate is not derived or when the candidate is not available.

[1002] For example, the list may include an MPM list, a TIMD candidate list, a TMRL intra prediction mode candidate list, and an intra prediction mode (IPM) list.

[1003] For example, the derived mode can be used in the prediction methods described in the embodiments. The prediction method can be a prediction method that constructs an IPM list.

[1004] For example, the derived mode can be used as an intra prediction mode used in a prediction method.

[1005] For example, the derived mode can be used in prediction methods such as GPM, SGPM, and Intra Block Copy with Geometry Partitioning mode (IBC-GPM). The derived mode can be used as a replacement for a specific intra prediction mode used in these prediction methods.

[1006] In one embodiment, the derived modes can be used to replace and / or extend an intra prediction mode derived from a block; an intra prediction mode of the block; or an intra prediction mode provided from the block, when the surrounding blocks are encoded / decoded using a prediction technique that does not store information of directional intra prediction modes.

[1007] For example, prediction techniques that do not store information of directional intra prediction modes may include prediction techniques described in the embodiments, Matrix-based Intra Prediction (MIP), Matrix-based Position-Dependent Intra Prediction (MPDIP or Position Dependent Prediction (PDP), IntraTMP, IBC, and EIP.

[1008] The derived modes can be used to replace and / or extend the intra prediction mode of a block, if the block has a specific intra prediction mode. The specific intra prediction mode and the replaced intra prediction mode can be directional intra prediction modes.

[1009] For example, if the intra prediction mode applied to the prediction of the surrounding blocks and the derived mode are different, when constructing a list of target blocks, the derived mode can replace the directional intra prediction mode applied to the prediction of the surrounding blocks.

[1010] For example, when the intra prediction mode applied to the prediction of the surrounding blocks and the derived mode are different, when constructing the list of target blocks, one or more of the derived N modes may be used / added as a new candidate to the list.

[1011] For example, if the intra prediction mode and the derived mode applied to the prediction of the surrounding block are different from each other, one of the prediction mode and the derived mode actually applied to the prediction of the surrounding block can be selected based on information of the surrounding samples.

[1012] The derived modes can be used to replace and / or extend the intra prediction mode of a block, if the block has a specific intra prediction mode. The specific intra prediction mode and the replaced intra prediction mode can be directional intra prediction modes.

[1013] When constructing the list described in the embodiments, one or more of the derived N modes may be used / added as new candidates to the list.

[1014] Whether the derived mode is used in the prediction described in the embodiments may be determined by whether a non-directional intra prediction mode is used in the intra prediction process for the block.

[1015] For example, if a block is reconstructed using a non-directional intra prediction mode during the intra prediction process for the block, the derived mode may not be used in the prediction described in the embodiments.

[1016] For example, if a block is reconstructed using a non-directional intra prediction mode during the intra prediction process for the block, the derived mode can be used in the prediction described in the embodiments.

[1017]

[1018] Mapping between template matching prediction and intra prediction modes

[1019] Below, embodiments are described that use template matching prediction for image encoding / decoding; and template matching block vectors determined through template matching prediction;

[1020] In embodiments, template matching prediction may include an intra-template matching mode, such as IntraTMP.

[1021] For example, a picture or block can be encoded / decoded using one of intra prediction, inter prediction, template matching prediction, template matching prediction fusion mode, template matching prediction local illumination compensation (LIC) mode, template matching sub-pel prediction mode, template matching linear filtering mode, and IBC mode.

[1022] Template matching prediction can search for matching regions within a search region that most closely resemble the template of the target block. Once a matching region is found, the corresponding region determined by the matching region can be used as the target block and predicted block.

[1023] The template may be part of the restoration area surrounding the target block.

[1024] The search area may be all or part of the restoration area within the target picture that contains the target block.

[1025] The matching area may be the area with the minimum error cost compared to the template. The shape of the template and the shape of the matching area may be identical.

[1026] The positional relationship between the template and target block may be identical to the positional relationship between the matching area and the corresponding area. In other words, the corresponding area may be the area into which the target block is moved according to a block vector. The block vector may be a vector from the template to the matching area.

[1027] A block vector derived by template matching prediction may be referred to as a template matching block vector.

[1028] The surrounding restoration area may be a portion of a plurality of spatial areas adjacent to the target block. The plurality of spatial areas may include an upper area located at the top of the target block, an upper left area located at the upper left of the target block, a left area located at the left side of the target block, and a lower left area located at the lower left side of the target block.

[1029] In one embodiment, the search area may be derived as described below based on the horizontal and vertical lengths of the target block.

[1030] For example, the navigation area may be located in the upper left, upper right, or left of a specific location. The navigation areas may be located in the upper left, upper right, and left of a specific location, respectively.

[1031] For example, the specific location may be the upper left location of the CTB containing the target block.

[1032] The search area can be a rectangular area of ​​a specific size.

[1033] The horizontal length of the search area can be M times the horizontal length of the target block. The vertical length of the search area can be N times the vertical length of the target block. Each of M and N can be an integer greater than or equal to 1. For example, M or N can be 6.

[1034] For example, the search area can be limited to the range of block vectors that can occur in IBC mode. In other words, areas / locations that cannot be pointed to by block vectors can be excluded from the search area.

[1035] For example, the search area may further include a restored area within the CTB containing the target block, in addition to the area composed of rectangular areas.

[1036] For example, the search area can be derived using the coding parameters described in the embodiments.

[1037] An already encoded / decoded area within the target picture may be a reconstructed area prior to application of at least one of the specific in-loop filterings described in the embodiments.

[1038] In embodiments, specific in-loop filtering may include chroma scaling, luma mapping, deblocking filtering, adaptive sample offset, and adaptive in-loop filtering.

[1039] Additionally, the already encoded / decoded area within the target picture may be a reconstructed area on which at least one of the specific in-loop filterings described in the embodiments has been performed.

[1040] The restoration region may be a region constructed using at least one of the prediction samples and the residual samples; or a combination of the prediction samples and the residual samples.

[1041] A prediction sample may mean a sample encoded / decoded using at least one of the prediction methods described in the embodiments, such as intra prediction, inter prediction, IBC, and template matching prediction.

[1042] For example, if the restoration region is a restoration region composed of prediction samples, the restoration region may not include regions / samples generated by inter prediction.

[1043] For example, if the restoration region is a restoration region composed of prediction samples, the region / sample encoded / decoded by inter prediction can be replaced with a specific value.

[1044] For example, a particular value may be a value generated by padding using prediction samples surrounding the sample being replaced.

[1045] For example, the specific value may be a value calculated using statistical values ​​of prediction samples surrounding the sample being replaced and coding parameters of the prediction samples surrounding the sample being replaced.

[1046] For example, the specific value may be a value calculated by filtering, sampling, and interpolation described in the embodiments.

[1047] The encoding / decoding method using template matching prediction described in the embodiments can be used in at least one of the following cases: when the luma component and the chroma component each have independent block partitioning structures (i.e., when a dual tree is used); and when the luma component and the chroma component have the same block partitioning structure (i.e., when a single tree is used).

[1048] The matching block with the smallest error cost derived through template matching of the embodiments can be used as a reference block.

[1049] Intra prediction modes can be mapped using information from reference blocks.

[1050] In one embodiment, when a target block or a neighboring block uses prediction using the IBC mode or template matching prediction mode, the block may be encoded / decoded by at least one of the methods described below. The target block may be a luma component block.

[1051] For example, when a list for a luma component is constructed, if the template matching prediction mode or the IBC mode is used for the surrounding blocks of the target block, the information required for encoding / decoding the target block can be derived using the block vectors of the surrounding blocks and the positions of the surrounding blocks. For example, information indicated by the block vectors of the surrounding blocks and / or the positions of the surrounding blocks can be mapped to the information required for encoding / decoding the target block. For example, by applying the block vectors to the positions of the surrounding blocks, the information of the samples / blocks at the positions indicated can be mapped to the information required for encoding / decoding the target block.

[1052] In embodiments, a location indicated by applying a specific block vector to a specific location may be a location indicated by the sum of the specific location and the block vector.

[1053] For example, the list may include lists described in the embodiments, such as an MPM list and an Intra Prediction Mode (IPM) list.

[1054] For example, information required for encoding / decoding a target block may include an intra prediction mode mapped using a block vector.

[1055] For example, if the luma component and the chroma component each have independent block division structures, and the IBC mode or the template matching prediction mode is used for the target chroma component block, information required for encoding / decoding the target chroma component block can be derived using coding information of the luma component block corresponding to the target chroma component block. For example, the coding information of the luma component block can be mapped to information required for encoding / decoding the target chroma component block.

[1056] For example, information required for encoding / decoding a target chroma component block may include a mapping mode of a luma component.

[1057] The encoding information may include coding parameters described in the embodiments.

[1058] When prediction using the IBC mode or template matching prediction mode is used for the target luma component block, a specific intra prediction mode as described below may be mapped to the target luma component block. Here, the mapped intra prediction mode may be referred to as a template matching prediction mode.

[1059] Template matching prediction modes can be mapped using coding information of a block (e.g., CU) containing a specific sample.

[1060] In one embodiment, a specific sample may be a sample at one or more of the sample locations described below among the samples included in the matching block. Samples at one or more of the sample locations described below among the samples included in the matching block may be selected as candidate samples, and a specific sample may be determined among the candidate samples.

[1061]

[1062] Figure 17 shows the locations of samples within a reference block according to an example.

[1063] A specific intra prediction mode can be mapped using coding information of a block including at least one of the upper left sample A, the upper right sample B, the center sample C, the lower left sample D, and the lower right sample E of the reference block.

[1064] In embodiments, the central sample C may be a sample located at the center of the reference block, a sample adjacent to the upper left of the center of the reference block, a sample adjacent to the upper right of the center of the reference block, a sample adjacent to the lower left of the center of the reference block, or a sample adjacent to the lower right of the center of the reference block.

[1065] A specific intra prediction mode can be mapped using the coding information of a block containing such samples. The block can be a CU.

[1066] If an intra prediction mode cannot be derived from a sample or block, an intra prediction mode described in embodiments, such as a planar mode or a DC mode, can be derived as a specific intra prediction mode.

[1067]

[1068] Figure 18 illustrates a mapping of information using the location of a target block and the block vector of an adjacent block according to an example.

[1069] Figure 19 illustrates a mapping of information using the location of adjacent blocks and the block vector of the adjacent blocks according to an example.

[1070] In one embodiment, M intra prediction modes can be mapped using block vectors of encoded / decoded blocks using template matching prediction modes.

[1071] Here, M can be an integer greater than or equal to 1. For example, M can be 4.

[1072] Here, N block vectors can be used for encoding / decoding the target block. N can be an integer greater than or equal to 1.

[1073] For example, a specific sample position or block of a reference block may be given a priority. Once sample positions with priorities are configured, the availability of an intra prediction block can be determined using the coding information of the luma component block at the specific sample position for mapping the intra prediction mode.

[1074] For example, candidates for a particular sample or block can be constructed using A, B, C, D and E described above with reference to FIG. 17.

[1075] For example, intra prediction modes of A, B, C, D, and E can be used as candidates. Alternatively, intra prediction modes derived using coding information of A, B, C, D, and E can be used as candidates.

[1076] Here, it can be determined whether an intra prediction mode is available at a position with a high priority. If an intra prediction mode is available at a position with a high priority, the candidate at the position can be mapped to the intra prediction mode of the target block.

[1077] At this time, when mapping intra prediction modes, a higher priority may be given to the directional intra prediction mode.

[1078] For example, as illustrated in FIG. 18, when a list for encoding / decoding a luma block is constructed, if the surrounding blocks of the target block use the template matching prediction mode or the IBC mode, information required for encoding / decoding the target block can be derived using the block vectors of the surrounding blocks and the positions of the target block. For example, information indicated by the block vectors of the surrounding blocks and / or the positions of the target block can be mapped to information required for encoding / decoding the target block. For example, information of a sample / block indicated by applying the block vectors of the surrounding blocks to the positions of the target block can be mapped to information required for encoding / decoding the target block.

[1079] For example, as illustrated in FIG. 19, when a list for encoding / decoding a luma block is constructed, if a neighboring block of a target block uses a template matching prediction mode or an IBC mode, information required for encoding / decoding the target block can be derived using a block vector of the neighboring block and a position of the neighboring block. For example, information indicated by the block vector of the neighboring block and / or the position of the neighboring block can be mapped to information required for encoding / decoding the target block. For example, by applying a block vector to the position of the neighboring block, information of a sample / block at the indicated position can be mapped to information required for encoding / decoding the target block.

[1080] Here, the list can be an MPM list or an IPM list.

[1081] Here, the information required for encoding / decoding the target block may include an intra prediction mode.

[1082] Here, the location of the target block may be one or more of the locations of A, B, C, D and E described above with reference to FIG. 17 within the target block.

[1083] Here, the location of the surrounding block may be one or more of the locations of A0, A1, B0, B1, and B2 shown in FIGS. 18 and 19.

[1084] Here, the location where the block vector is applied may be one or more of the locations A, B, C, D, and E described above with reference to FIG. 17 within the target block or surrounding blocks.

[1085] Here, scaling can be applied to the block vector when the block vector is applied to a specific location of the target block.

[1086] In embodiments, the MPM list may include a primary MPM (PMPM) and a secondary MPM (SMPM).

[1087] When the PMPM list is constructed, the L candidates in the PMPM list can be reordered according to the template costs of the candidates.

[1088] For example, when a list for luma block encoding / decoding is constructed, if the surrounding blocks of the target block use the template matching prediction mode or the IBC mode, a method for deriving an intra prediction mode by using analysis of the reference template (e.g., DIMD) can be applied to the surrounding blocks.

[1089] Here, the list can include an MPM list and an IPM list.

[1090] Here, the intra prediction mode derived using the analysis of the reference template can be applied to the entire area of ​​the surrounding block, and can be applied to a location sampled from the area of ​​the surrounding block.

[1091] Here, the surrounding blocks may be already encoded / decoded blocks.

[1092] In one embodiment, the method of performing mapping of intra prediction modes using information of template matching modes described in the embodiments can be applied to template matching fusion mode.

[1093] Template matching prediction modes can be mapped using coding information of a block (e.g., CU) containing a specific sample.

[1094] In one embodiment, a specific sample may be a sample at one or more of the sample locations described below among the samples included in the matching block. Samples at one or more of the sample locations described below among the samples included in the matching block may be selected as candidate samples, and a specific sample may be determined among the candidate samples.

[1095] As illustrated in FIG. 17, a specific intra prediction mode can be mapped using coding information of a block including at least one of the upper left sample A, the upper right sample B, the center sample C, the lower left sample D, and the lower right sample E of the reference block.

[1096] A specific intra prediction mode can be mapped using the coding information of a block containing such samples. The block can be a CU.

[1097] If an intra prediction mode cannot be derived from a sample or block, an intra prediction mode described in embodiments, such as a planar mode or a DC mode, can be derived as a specific intra prediction mode.

[1098] In one embodiment, M intra prediction modes can be mapped using block vectors of encoded / decoded blocks using template matching prediction modes.

[1099] Here, M can be an integer greater than or equal to 1. For example, M can be 4.

[1100] Here, N block vectors can be used for encoding / decoding the target block. N can be an integer greater than or equal to 1.

[1101] For example, a specific sample position or block of a reference block may be given a priority. Once sample positions with priorities are configured, the availability of an intra prediction block can be determined using the coding information of the luma component block at the specific sample position for mapping the intra prediction mode.

[1102] For example, candidates for a particular sample or block can be constructed using A, B, C, D and E described above with reference to FIG. 17.

[1103] For example, intra prediction modes of A, B, C, D, and E can be used as candidates. Alternatively, intra prediction modes derived using coding information of A, B, C, D, and E can be used as candidates.

[1104] Here, it can be determined whether an intra prediction mode is available at a position with a high priority. If an intra prediction mode is available at a position with a high priority, the candidate at the position can be mapped to the intra prediction mode of the target block.

[1105] At this time, when mapping intra prediction modes, a higher priority may be given to the directional intra prediction mode.

[1106] For mapping of template matching prediction mode, coding information of a block including at least one sample among samples corresponding to multiple matching blocks may be used.

[1107] For example, when a list for encoding / decoding a luma block is constructed, if the surrounding blocks of the target block use the template matching prediction fusion mode, information required for encoding / decoding the target block can be derived using the block vectors of the surrounding blocks and the positions of the surrounding blocks. For example, information indicated by the block vectors of the surrounding blocks and / or the positions of the surrounding blocks can be mapped to information required for encoding / decoding the target block. For example, by applying the block vectors to the positions of the surrounding blocks, information of the samples / blocks at the positions indicated can be mapped to information required for encoding / decoding the target block.

[1108] Here, the list can be an MPM list or an IPM list.

[1109] Here, the information required for encoding / decoding the target block may include an intra prediction mode.

[1110] Here, the location of the surrounding block may be one or more of the locations of A0, A1, B0, B1, and B2 shown in FIGS. 18 and 19.

[1111] Here, the location where the block vector is applied may be one or more of the locations A, B, C, D, and E described above with reference to FIG. 17 within the target block or surrounding blocks.

[1112] In embodiments, the MPM list may include a primary MPM (PMPM) and a secondary MPM (SMPM).

[1113] When the PMPM list is constructed, the L candidates in the PMPM list can be reordered according to their template costs. The template cost can be derived in a manner similar to TIMD, and the template can be a single reference sample line adjacent to the current block.

[1114] For example, when a list is constructed for luma block encoding, if the surrounding blocks of the target block use the template matching prediction mode or the IBC mode, a method of deriving an intra prediction mode by using analysis of the reference template (e.g., DIMD) can be applied to the surrounding blocks.

[1115] Here, the list can include an MPM list and an IPM list.

[1116] Here, the intra prediction mode derived using the analysis of the reference template can be applied to the entire area of ​​the surrounding block, and can be applied to a location sampled from the area of ​​the surrounding block.

[1117] Here, the surrounding blocks may be already encoded / decoded blocks.

[1118] Information of the intra prediction mode derived by the method described above in the embodiments can be stored as information required for encoding / decoding of the target block.

[1119] For example, the stored intra prediction mode can be used in the process of transforming / inverse transforming a target block encoded / decoded in template matching prediction mode.

[1120] Here, a stored intra prediction mode may be used by applying an intra prediction mode derived by using analysis on a reference template; or an intra prediction mode stored by using block vector information.

[1121] For example, the stored intra prediction mode can be used as a candidate for encoding / decoding the chroma components described in the embodiments. Here, the candidate can be a candidate from the list described in the embodiments.

[1122]

[1123] Template Matching Prediction Merge Technology

[1124] Below, embodiments of a template matching prediction method and a template matching prediction merging method can be described.

[1125] Template matching prediction methods may include the methods listed below:

[1126] - Template matching prediction method using one block vector;

[1127] - Template matching fusion method using one or more block vectors;

[1128] - Template matching prediction block illumination compensation method;

[1129] - Template matching block vector compensation method;

[1130] - Template matching prediction merging method; and

[1131] - How to use template matching described in the examples.

[1132] When the template matching prediction mode is used for the target block, entropy encoding / decoding for the template matching prediction mode information can be used.

[1133] For example, information about template matching prediction may include at least one of the information listed below:

[1134] - An indicator indicating whether template matching prediction is used;

[1135] - An index indicating the block vector used for template matching prediction;

[1136] - A directive indicating template matching fusion mode;

[1137] - An index indicating a candidate group for template matching fusion mode;

[1138] - An indicator indicating the number of block vectors for template matching fusion mode;

[1139] - An index indicating the number of block vectors for template matching fusion mode;

[1140] - Multiple indices indicating candidate block vectors for template matching fusion mode;

[1141] - An index that indicates how to apply weights for the template matching fusion mode;

[1142] - A directive that instructs the roughness compensation of the template matching prediction block;

[1143] - An index indicating the method of compensation for the roughness of the template matching prediction block;

[1144] - A directive indicating a template matching prediction block vector compensation method; and

[1145] - An index that indicates how to apply the template matching prediction block vector compensation method.

[1146] Information about intra prediction described in the embodiments can be signaled based on the coding parameters described in the embodiments.

[1147] For example, information about intra prediction may not be signaled based on one or more of the slice type, the block size, and the block shape.

[1148] Here, information about unsignaled intra prediction can be derived using the coding parameters described in the embodiments.

[1149] For example, if the target block has a predefined size, information regarding template-matching prediction for the target block may not be signaled. Here, information regarding template-matching prediction for a previously encoded / decoded upper block may be used to replace the unsigned information. Here, template-matching prediction for the upper block may mean template-matching prediction for the size of the upper block.

[1150] For example, if the target block is rectangular in shape, information regarding template matching prediction for the target block may not be signaled. Information regarding template matching prediction for previously encoded / decoded upper blocks may be used to replace the unsignaled information.

[1151] When template matching prediction is performed for a target block, it can be derived whether a template matching prediction merging method is performed using information of restoration blocks surrounding the target block.

[1152] For example, when template matching prediction is applied to restoration blocks surrounding a target block, a merged template matching-related method can be derived by merging template matching-related methods applied to the restoration blocks, and the merged template matching-related method can be used as a candidate for the final prediction method.

[1153] For example, when template matching prediction is applied to restoration samples around a target block, a merged template matching-related method can be derived by merging template matching-related methods applied to restoration blocks including restoration samples, and the merged template matching-related method can be used as a candidate for the final prediction method.

[1154] For example, the target block may include one or more of the following regions: a surrounding block of the target block, a region vertically adjacent to the top of the target block, a region horizontally adjacent to the top of the target block, a region diagonally adjacent to the top right of the target block, a region diagonally adjacent to the top left of the target block, a region diagonally adjacent to the bottom left of the target block, and a region not adjacent to the target block. Alternatively, the surrounding block of the target block may include one of these regions.

[1155] When a template matching merging method is performed on a target block, information on modes used for template matching prediction can be derived using information on a restoration area surrounding the target block.

[1156] For example, if template matching prediction is applied to restoration blocks surrounding the target block, template matching related methods applied to the restoration blocks can be used without additional signaling.

[1157] For example, if template matching prediction is applied to restoration samples around the target block, template matching related methods applied to the restoration block containing the restoration samples can be used without additional signaling.

[1158] For example, the target block may include one or more of the following regions: a surrounding block of the target block, a region vertically adjacent to the top of the target block, a region horizontally adjacent to the top of the target block, a region diagonally adjacent to the top right of the target block, a region diagonally adjacent to the top left of the target block, a region diagonally adjacent to the bottom left of the target block, and a region not adjacent to the target block. Alternatively, the surrounding block of the target block may include one of these regions.

[1159] Whether a template matching prediction merging method indicator for a target block is signaled can be determined based on information in the restoration area surrounding the target block.

[1160] Here, information in the restoration area may include coding parameters of objects (e.g., blocks or pictures) within the restoration area.

[1161] In one embodiment, whether a template matching prediction merging method indicator for a target block is signaled may be determined based on the state of other indicators described in the embodiments.

[1162] In the embodiments, when a specific indicator indicates whether a specific function is performed, a value of 1 or true of the specific indicator may mean that the specific function is performed. A value of 0 or false of the specific indicator may mean that the specific function is not performed. That is, in the embodiments, "when the value of the specific indicator is 1 or true" may mean "when the specific indicator indicates that the specific function is performed." "When the value of the specific indicator is 0 or false" may mean "when the specific indicator indicates that the specific function is not performed."

[1163] For example, if the value of the directive indicating the template matching fusion mode of the target block is a predefined value, the template matching prediction merging method directive may not be signaled for the target block. For example, the predefined value may be 1.

[1164] For example, if the value of the indicator indicating the number of block vectors for the template matching fusion mode of the target block is a predefined value, the template matching prediction merging method indicator may not be signaled for the target block. For example, the predefined value may be 1.

[1165] For example, if the value of the indicator indicating the illumination compensation of the template matching prediction block of the target block is a predefined value, the template matching prediction merging technology indicator may not be signaled for the target block. For example, the predefined value may be 1.

[1166] For example, if the directive indicating the template matching prediction block vector compensation method of the target block is a predefined value, the template matching prediction merging method directive may not be signaled for the target block. For example, the predefined value may be 1.

[1167] In one embodiment, whether other directives described in the embodiments are signaled may be determined based on the state of the template matching prediction merging method directive for the target block.

[1168] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the indicator indicating the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1169] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the index indicating the candidate group for the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1170] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the indicator indicating the number of block vectors for the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1171] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the index indicating the number of block vectors for the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1172] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, multiple indices indicating candidate block vectors for the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1173] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the index indicating the weighting method for the template matching fusion mode may not be signaled. For example, the predefined value may be 1.

[1174] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the indicator indicating the roughness compensation of the template matching prediction block may not be signaled. For example, the predefined value may be 1.

[1175] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the index indicating the roughness compensation method of the template matching prediction block may not be signaled. For example, the predefined value may be 1.

[1176] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the indicator indicating the template matching prediction block vector compensation method may not be signaled. For example, the predefined value may be 1.

[1177] For example, if the value of the template matching prediction merging method indicator for the target block is a predefined value, the index indicating how to apply the template matching prediction block vector compensation method may not be signaled. For example, the predefined value may be 1.

[1178]

[1179] Intra prediction using analysis of surrounding blocks of chroma components

[1180] Below, embodiments using image encoding / decoding using an intra prediction mode of a chroma component; and a mode in which a method using an intra prediction mode of a chroma component is applied; are described.

[1181] For example, the intra prediction mode of the chroma component can be encoded / decoded using one of a directional intra prediction mode; a non-directional intra prediction mode (e.g., a DC mode or a planar mode); an intra prediction mode derived using analysis of decoded samples (e.g., an intra prediction mode derived using DIMD chroma); a luma component intra prediction mode at a position corresponding to a position of the chroma component (e.g., an intra prediction mode derived using Direct Mode (DM)); a chroma component fusion mode; and a chroma component cross-component prediction (CCP) mode.

[1182] Here, an intra prediction mode that does not include an inter-component prediction mode can be referred to as a chroma non-inter-component prediction mode.

[1183] Additionally, a mode may be included in which a linear model or a model using a plus (+) shaped filter model is applied to the relationship between chroma components and luma components as a chroma component-to-luma component prediction (CCP) mode.

[1184] Here, the filter model may have a different shape than described in the embodiments.

[1185] For example, a linear model of a luma component and a chroma component can be constructed, and the constructed model can be used for chroma component block prediction.

[1186] Here, there can be M linear models. M can be an integer greater than or equal to 1.

[1187] For example, a plus (+) shaped filter model for the relationship between a luma component and a chroma component can be constructed, and the constructed filter model can be used for chroma component block prediction.

[1188] Here, there can be M linear models. M can be an integer greater than or equal to 1.

[1189] The chroma non-component-to-chroma prediction mode, which utilizes analysis of encoded / decoded samples of chroma components, can extract orientation and magnitude (or size) information by analyzing vertical and horizontal gradients between samples of a target block's template and surrounding samples. The extracted information can be used to generate a histogram of gradients (HoG).

[1190] In embodiments, HoG may mean a histogram of oriented gradient.

[1191] Here, the template may include already encoded / decoded samples surrounding the target block. The template may use N reference sample lines. The number of reference sample lines may be N. N may be an integer greater than or equal to 1.

[1192] Here, the chroma non-component intra prediction mode can be derived using HoG.

[1193] For example, the chroma non-component intra prediction modes may be the L intra prediction modes with the largest accumulated magnitude values ​​in the HoG.

[1194] Here, L can be an integer greater than or equal to 1. For example, L can be 5.

[1195] Here, as samples for analysis, a template of a luma component block and a template of a chroma component block at a corresponding position corresponding to the position of the target block can be used.

[1196] In embodiments, the corresponding position may mean a position corresponding to the position of the target block. The corresponding position may mean a position identical to the position of the target block.

[1197] In embodiments, the corresponding location of a specific object / location may mean a location corresponding to the specific object / location. The corresponding location of a specific object / location may mean a location identical to the location of the specific object / location.

[1198] Here, the templates of the luma component blocks at the corresponding positions may include a top template, a top left template, and a bottom left template.

[1199] In one embodiment, when a template of a luma component block at a particular location is not available, a restored luma component block can be used for analysis.

[1200] For example, when the top template is not available, a portion of the top of the restored luma component block can be used for analysis.

[1201] For example, when the template in the lower left corner is not available, a portion of the lower left corner of the restored luma component can be used for analysis.

[1202] In one embodiment, the N intra prediction modes described above can be derived, and the derived intra prediction modes can be used to generate a chroma component prediction block.

[1203] For example, one intra prediction mode with the highest cumulative magnitude value can be used as the chroma component intra prediction mode.

[1204] For example, a weighted fusion of L intra prediction modes with the highest cumulative magnitude values ​​can be used as the chroma component intra prediction mode.

[1205] For example, a weighted fusion of L intra prediction modes with the highest cumulative magnitude values ​​and non-directional modes (e.g., DC mode and planar mode) can be used as a chroma component intra prediction mode.

[1206] For example, a weighted fusion of one intra prediction mode with the highest cumulative magnitude value and a non-directional mode (e.g., a DC mode and a planar mode) can be used as the chroma component intra prediction mode.

[1207] For example, a weighted fusion of L intra prediction modes having the highest cumulative magnitude values ​​and a value derived by applying down-sampling to the luma component at the corresponding location can be used as the chroma component intra prediction mode.

[1208] For example, a weighted fusion of one intra prediction mode having the highest cumulative magnitude value and a value derived by down-sampling the luma component at the corresponding location can be used as the chroma component intra prediction mode.

[1209]

[1210] Intra prediction mode of chroma components based on fusion

[1211] Figure 20 shows the surrounding blocks of another target block as an example.

[1212] Figure 21 illustrates adjacent blocks and non-adjacent blocks of a target block according to an example.

[1213] In FIG. 20 and FIG. 21, the locations of pixels and blocks from which information required to determine the weights of the chroma component fusion mode is derived can be exemplified.

[1214] In addition to the locations illustrated in FIGS. 20 and 21, pixels or blocks described as referenced in the embodiments may be used to determine the weights of the chroma component fusion mode.

[1215] The chroma component fusion mode of the embodiment may use a weighted fusion of a chroma non-component-to-component prediction mode; an intra prediction mode using analysis of encoded / decoded samples; or a chroma component-to-component prediction mode.

[1216] The weights of the chroma component fusion mode can be determined using information used for encoding / decoding of neighboring blocks of the target block, or non-adjacent blocks, as illustrated in FIGS. 20 and 21.

[1217] For example, the information used for encoding / decoding may include chroma component-to-component prediction modes, chroma non-component-to-component prediction modes, and coding parameters described in the embodiments.

[1218] Here, information of the surrounding block can be derived using information of a block containing M pixels or the above pixels. The block can be a CU. M can be an integer greater than or equal to 2.

[1219] For example, the weights described in the embodiments can be determined using information on the chroma component-to-chroma prediction mode of the surrounding blocks; and the chroma non-component-to-chroma prediction mode.

[1220] For example, the weights described in the embodiments can be determined using information on the chroma component prediction mode of the surrounding block; and the size of the surrounding block.

[1221] The weight can be determined using the minimum value of the width and height of the surrounding blocks.

[1222] The weight can be determined using the maximum value of the width and height of the surrounding blocks.

[1223] The weights can be determined using the area of ​​the surrounding blocks.

[1224] For example, weights can be determined using intra prediction information of surrounding blocks of a luma component prediction block at a corresponding location.

[1225] For example, weights can be determined using information about the chroma component-to-chroma prediction mode and the chroma non-component-to-chroma prediction mode of non-adjacent blocks.

[1226] For example, weights can be determined using the chroma component prediction mode of a non-adjacent block and the size information of the above block.

[1227] For example, the weights can be determined using the template cost of the chroma component prediction mode of the surrounding blocks.

[1228] The weight can be determined using the minimum value of the width and height of the surrounding blocks.

[1229] The weight can be determined using the maximum value of the width and height of the surrounding blocks.

[1230] The weights can be determined using the area of ​​the surrounding blocks.

[1231] For example, the weights can be determined using the difference between the values ​​derived by down-sampling the luma component prediction block at the corresponding location; and the chroma component prediction block.

[1232] In one embodiment, a weighted fusion of N chroma component prediction modes may be used as the chroma component fusion mode. N may be an integer greater than or equal to 2.

[1233] For example, weighted fusion of chroma non-component-to-chroma prediction mode and chroma component-to-chroma prediction mode can be used.

[1234] For example, weighted fusion of chroma non-component-to-component prediction mode and chroma non-component-to-component prediction mode can be used.

[1235] For example, weighted fusion of chroma component-to-chroma prediction modes and chroma component-to-chroma prediction modes can be used.

[1236] For example, weighted fusion of intra prediction mode and chroma non-component inter prediction mode using analysis on encoded / decoded samples can be used.

[1237] For example, weighted fusion of intra prediction mode and inter-chroma component prediction mode using analysis of encoded / decoded samples can be used.

[1238] For example, one of the intra prediction modes described above; and a weighted fusion of chroma non-directional modes (e.g., DC and planar modes) may be used.

[1239] The encoding / decoding method for the chroma component fusion mode described in the embodiments can be used in at least one of the following cases: when the luma component and the chroma component each have independent block partitioning structures (i.e., when a dual tree is used); and when the luma component and the chroma component have the same block partitioning structure (i.e., when a single tree is used).

[1240]

[1241] Derivation of intra prediction mode using reference template / sample analysis and multiple reference lines

[1242] An intra prediction mode can be derived using analysis of a reference template / sample. The derived intra prediction mode can be used as a candidate in a candidate list and as the intra prediction mode of a target block.

[1243] The template may be a restored area around the target block.

[1244] For example, analysis of the reference sample may include decoder-side intra-mode derivation (DIMD).

[1245] Analysis of reference samples of the target block can be applied in the process of deriving the intra prediction mode.

[1246] Analysis of the reference samples of the target block may be an analysis of the directionality of the pre-encoded / decoded reference samples with respect to the template of the target block.

[1247] This analysis can extract orientation and magnitude (or size) information by analyzing vertical and horizontal gradients between a target sample and surrounding samples. The extracted information can be used to generate a histogram of gradients (HoG).

[1248] In embodiments, HoG may mean a histogram of oriented gradient.

[1249] For example, in a HoG, bins can be generated, each corresponding to an available intra prediction mode. A magnitude of each bin can be generated. The magnitude of a bin can represent the magnitude of the intra prediction mode corresponding to the bin. The available intra prediction modes can include directional intra prediction modes.

[1250] To generate the HoG of the embodiments, other information described in the embodiments, such as coding parameters, may be used.

[1251] Additionally, information about the HoG, such as orientation and magnitude, and information used to generate the HoG, may be determined using specific information about the samples of the embodiments. For example, the specific information may include geometrical characteristics of the samples.

[1252] For this analysis, a reference sample line L other than reference sample line 0 may be used. L may be an integer greater than or equal to 1. The reference sample line L may be a reference sample line other than reference sample line 0 among the multiple reference sample lines described with reference to FIG. 3.

[1253] Here, the reference sample line L may be referred to as a multiple reference sample line.

[1254] Multiple reference sample lines can be used for analysis.

[1255] For example, reference sample line L, reference sample line L+1 and reference sample line L+2 can be used together for analysis.

[1256] Multiple reference sample lines can be used independently of each other.

[1257] N intra prediction modes can be derived using reference template analysis, where N can be an integer greater than or equal to 2.

[1258] Each derived intra prediction mode can be determined using one or more coding parameters.

[1259] N intra prediction modes can be derived using multiple reference lines.

[1260] In one embodiment, the N modes induced by the analysis can use the reference sample line M as a template.

[1261] Prediction for a reference sample line M can be performed using the N modes derived by the analysis.

[1262] For example, prediction can be performed for N intra prediction modes using reference samples of reference sample line M+1. Here, a specific intra prediction mode among the N intra prediction modes can be selected based on an error between a value derived by the prediction and a value of a reference sample of reference sample line M.

[1263] Here, one or more reference sample lines can be used.

[1264] For example, M can be an integer greater than or equal to 0. Reference sample line 0 can mean the reference sample line closest to the target block.

[1265] Here, the N intra prediction modes derived may be intra prediction modes obtained by multiple reference sample lines.

[1266] Prediction for a template can be performed using a reference sample line that is outside the normal reference sample line, such as template line M+1. The template line may refer to a reference sample line or a sample line used as a template.

[1267] For example, when performing prediction for template line 2 of a target block, reference sample line 3 or reference sample line 4 can be used as a reference sample for the template line.

[1268] Here, an intra prediction mode can be derived using the error between the values ​​of the predicted template and the values ​​of the restored template. The restored template may be a template composed of restored samples.

[1269] In embodiments, the error between these templates may be referred to as template cost.

[1270] For example, the template cost can be calculated using a calculation method that utilizes the differences between pixels at the same location. Such calculation methods may include SAD, Sum of Absolute Transformed Differences (SATD), and Mean Removed - Sum of Absolute Differences (MR-SAD).

[1271] For example, template costs may be calculated using other cost / error calculation methods described in the embodiments.

[1272] The derived modes can be sorted in ascending order of the errors of the templates for the derived modes. The sorted intra prediction modes can be used.

[1273] In prediction of a block using analysis on a reference template / sample, a weighted fusion of multiple derived intra prediction modes can be used.

[1274] In embodiments, the derived intra prediction modes may be abbreviated as derived modes.

[1275] Here, weighted fusion can use modes and non-directional modes derived from analysis of the template.

[1276] Here, the mode induced by the analysis on the template may be a mode induced using multiple reference sample lines.

[1277] Here, the weighted fusion may be a mode induced by analysis of the template, or a mode induced using the multiple reference sample lines described above.

[1278] In one embodiment, the non-directional predictor may be a prediction block generated by at least one of the methods described below.

[1279] - The non-directional predictor may be one of the predictors using planar mode; DC mode; position-dependent prediction combination (PDPC); matrix-based intra prediction (MIP); matrix-based position-dependent intra prediction (MPDIP or Position Dependent Prediction; PDP); extrapolation filter-based intra prediction (EIP); or block vectors of neighboring blocks. In addition, the non-directional predictor may be information for prediction without directionality as described in the embodiments.

[1280] - One of the non-directional predictors can be selected based on the directionality of the induced mode or multiple induced modes.

[1281] - One of the non-directional predictors can be selected based on the shape of the template of the induced mode or multiple induced modes.

[1282] - The planner mode can be one of the normal planner mode, the planner horizontal mode, and the planner vertical mode.

[1283] Here, one of the aforementioned planar modes can be used as a reference for the directionality of the induced mode or multiple induced modes.

[1284] Here, one of the aforementioned planar modes can be used as a reference for the shape of the template of the induced mode or multiple induced modes.

[1285] In the process of weighted fusion, weights for the induced modes can be obtained by the methods described below.

[1286] For example, in the process of weighted fusion, weights for the induced modes can be obtained using the coding parameters of the embodiments.

[1287] For example, weights can be determined using magnitudes obtained using analysis on a reference template.

[1288] For example, weights can be determined using magnitudes obtained using analysis on a template using multiple reference lines.

[1289] In the process of weighted fusion, reference sample line 0 may be used, or reference sample line N may be used. N may be an integer greater than or equal to 1.

[1290] For example, left reference sample line 2 on the left side of the target block and top reference block 2 on the top side of the target block can be used.

[1291] In one embodiment, the weights used in the weighted fusion process may be derived using template costs. The template costs may be the difference between the predicted template and the reconstructed template.

[1292] The mode used to calculate the template cost may be a mode derived using multiple reference lines.

[1293] The weights used in weighted fusion can be derived using methods such as those described below.

[1294] - The weights can be obtained using the coding parameters of the embodiments.

[1295] - Weights can be derived in reverse order of template costs. ...

Claims

1. A step of deriving a prediction mode for a target block; A step of constructing a reference sample for prediction of the target block; and A step of performing the prediction for the target block using the reference sample and the derived prediction mode. A method for decrypting an image, comprising:

2. In paragraph 1, A candidate list for the prediction of the above target block is constructed, An image decoding method wherein the above candidate list is derived using a histogram of gradients (HoG).

3. In paragraph 2, An image decoding method wherein the above HoG is configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

4. In paragraph 2, An image decoding method, wherein the above HoG is configured based on the amount of change in pixel values ​​of predicted pixels of a prediction block.

5. In paragraph 1, An image decoding method, wherein an intra prediction mode derived using a restoration block or block vector of the target block is stored as an intra prediction mode of the target block.

6. In paragraph 5, A method for decoding an image, wherein the storage of the derived intra prediction mode is performed when the prediction does not use an intra prediction directional mode.

7. In paragraph 1, The above prediction is a method for decoding an image, wherein the intra prediction mode is derived by using analysis of a reference sample.

8. A step of deriving a prediction mode for the target block; A step of constructing a reference sample for prediction of the target block; and A step of performing the prediction for the target block using the reference sample and the derived prediction mode. A method for encoding an image, comprising:

9. In paragraph 8, A candidate list for the prediction of the above target block is constructed, An image encoding method wherein the above candidate list is derived using a histogram of gradients (HoG).

10. In paragraph 9, An image encoding method wherein the above HoG is configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

11. In paragraph 9, An image encoding method, wherein the above HoG is configured based on the amount of change in pixel values ​​of predicted pixels of a prediction block.

12. In paragraph 8, An image encoding method, wherein an intra prediction mode derived using a restoration block or block vector of the target block is stored as an intra prediction mode of the target block.

13. In paragraph 12, A method for encoding an image, wherein the storage of the derived intra prediction mode is performed when the prediction does not use an intra prediction directional mode.

14. In paragraph 8, The above prediction is a video encoding method that derives an intra prediction mode by using analysis of reference samples.

15. A computer-readable recording medium storing a bitstream generated by the image encoding method of Article 8.

16. In a computer-readable recording medium storing a bitstream, the bitstream comprises: Forecast information Including, Using the above prediction information, a prediction mode for the target block is derived, A reference sample for prediction of the above target block is constructed, A computer-readable recording medium for performing the prediction for the target block using the reference sample and the derived prediction mode.

17. In paragraph 16, A candidate list for the prediction of the above target block is constructed, A computer-readable recording medium in which the above candidate list is derived using a histogram of gradients (HoG).

18. In paragraph 17, A computer-readable recording medium wherein the above HoG is configured based on the amount of change in pixel values ​​of restored pixels within a specific area.

19. In paragraph 17, A computer-readable recording medium wherein the above HoG is configured based on the amount of change in pixel values ​​of predicted pixels of a prediction block.

20. In paragraph 16, A computer-readable recording medium in which an intra prediction mode derived using a restoration block or block vector of the target block is stored as an intra prediction mode of the target block.

Images