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

Abstract

Disclosed are a method, an apparatus, and a recording medium for image encoding / decoding. The method for image encoding / decoding comprises the steps of: determining a prediction mode for a chroma block, determining information for deriving a cross-component prediction model, performing prediction merging using the prediction model; and generating a chroma prediction block by using the cross-component prediction model. The chroma prediction block is derived by various prediction modes such as a cross-component linear model (CCLM), a gradient linear model (GLM), a filter-based linear model (FLM), and a convolutional cross-component model (CCCM).

Description

Technical Field

[0001] This disclosure generally relates to methods, apparatus, and storage media for image encoding / decoding. More specifically, this disclosure relates to methods, apparatus, and storage media for predicting chroma signals.

[0002] This application claims the benefit of Korean Patent Application No. 10-2023-0086006, filed on July 3, 2023, and Korean Patent Application No. 10-2024-0087712, filed on July 3, 2024, the entire contents of which are incorporated herein by reference. Background Technology

[0003] With the continued development of the information and communication industry, broadcast services supporting high-definition (HD) resolution have become widespread throughout the world. Through this widespread adoption, a large number of users have become accustomed to high-resolution and high-definition images and / or videos.

[0004] To meet user demand for high definition, numerous organizations have accelerated the development of next-generation imaging devices. In addition to High Definition TV (HDTV) and Full High Definition (FHD) TV, user interest in UHD TV has also increased, with UHD TV offering more than four times the resolution of Full High Definition (FHD) TV. With this growing interest, there is now a need for image encoding / decoding technologies specifically designed for images with higher resolution and greater definition.

[0005] As an image compression technique, there are various techniques (such as inter-frame prediction, intra-frame prediction, transform, quantization, and entropy coding).

[0006] Inter-frame prediction techniques are used to predict the values ​​of pixels included in the current frame using frames preceding and / or following the current frame. Intra-frame prediction techniques are used to predict the values ​​of pixels included in the current frame using information about the pixels in the current frame. Transform and quantization techniques can be used to compress the energy of residual signals. Entropy coding techniques are used to assign short codewords to frequently occurring values ​​and long codewords to less frequent values.

[0007] By utilizing these image compression techniques, data about images can be effectively compressed, sent, and stored. Summary of the Invention

[0008] Technical issues The embodiments aim to provide an apparatus, method, and storage medium for performing processing on a chromaticity signal by utilizing information from a luminance signal.

[0009] The embodiments are intended to provide an apparatus, method, and storage medium for performing predictions on chromaticity signals based on a cross-component approach.

[0010] Technical solution According to one aspect, an image decoding method is provided, comprising: determining a prediction method for chroma blocks; determining information for deriving a cross-component prediction model; and generating chroma prediction blocks using the cross-component prediction model.

[0011] The chromaticity prediction block can be generated based on a convolutional cross-component model that uses gradient and positional information.

[0012] The chromaticity prediction block can be generated based on a convolutional cross-component model using multiple downsampling filters.

[0013] The chromaticity prediction block can be generated based on a convolutional cross-component model using unsampled luminance samples.

[0014] Information from previously decoded blocks can be used to configure a reference region for cross-component prediction models.

[0015] The chromaticity prediction block can be generated based on cross-component prediction merging using a cross-component prediction model.

[0016] The chromaticity prediction block can be generated using prediction blocks generated by methods other than the cross-component prediction model.

[0017] According to another aspect, an image encoding method is provided, comprising: determining a prediction method for chroma blocks; determining information for deriving a cross-component prediction model; and generating chroma prediction blocks using the cross-component prediction model.

[0018] The chromaticity prediction block can be generated based on a convolutional cross-component model that uses gradient and positional information.

[0019] The chromaticity prediction block can be generated based on a convolutional cross-component model using multiple downsampling filters.

[0020] The chromaticity prediction block can be generated based on a convolutional cross-component model using unsampled luminance samples.

[0021] Information from previously encoded blocks can be used to configure a reference region for cross-component prediction models.

[0022] The chromaticity prediction block can be generated based on cross-component prediction merging using a cross-component prediction model.

[0023] The chromaticity prediction block can be generated using prediction blocks generated by methods other than the cross-component prediction model.

[0024] According to another aspect, a computer-readable storage medium is provided for storing a bitstream for image decoding, wherein the bitstream includes encoded information, a prediction method for chroma blocks is determined based on the encoded information, information for deriving a cross-component prediction model is determined, and the chroma prediction blocks are generated using the cross-component prediction model.

[0025] The chromaticity prediction block can be generated based on a convolutional cross-component model that uses gradient and positional information.

[0026] The chromaticity prediction block can be generated based on a convolutional cross-component model using multiple downsampling filters.

[0027] The chromaticity prediction block can be generated based on a convolutional cross-component model using unsampled luminance samples.

[0028] Information from previously decoded blocks can be used to configure a reference region for cross-component prediction models.

[0029] The chromaticity prediction block can be generated based on cross-component prediction merging using a cross-component prediction model.

[0030] Beneficial effects An apparatus, method, and storage medium are provided for performing processing on a chromaticity signal by utilizing information from a luminance signal.

[0031] An apparatus, method, and storage medium for performing predictions on chromaticity signals based on a cross-component approach are provided. Attached Figure Description

[0032] Figure 1 This is a block diagram illustrating the configuration of an embodiment of an encoding device to which the present disclosure is applied.

[0033] Figure 2 This is a block diagram illustrating the configuration of an embodiment of the decoding device to which this disclosure is applied.

[0034] Figure 3 It is a schematic diagram illustrating the partitioning structure of an image when it is encoded and decoded.

[0035] Figure 4 This is a diagram showing the form of prediction units (PUs) that a coding unit (CU) can include.

[0036] Figure 5 This is a diagram showing the form of a transformation unit (TU) that can be included in a CU.

[0037] Figure 6 This shows the block division based on the example.

[0038] Figure 7 This is a diagram illustrating an embodiment used to explain the intra-frame prediction process.

[0039] Figure 8 This is a diagram showing the reference samples used in the intra-frame prediction process.

[0040] Figure 9 This is a diagram illustrating an embodiment used to explain the inter-frame prediction process.

[0041] Figure 10 Spatial candidates according to an embodiment are shown.

[0042] Figure 11 The order in which motion information of spatial candidates is added to the merging list is shown according to an embodiment.

[0043] Figure 12 The transformation and quantization process is shown based on the example.

[0044] Figure 13 This shows a diagonal scan based on an example.

[0045] Figure 14 The horizontal scan is shown based on the example.

[0046] Figure 15 The vertical scan is shown according to the example.

[0047] Figure 16 This is a configuration diagram of an encoding device according to an embodiment.

[0048] Figure 17 This is a configuration diagram of a decoding device according to an embodiment.

[0049] Figure 18 This is a flowchart illustrating a target block prediction method and a bit stream generation method according to an embodiment.

[0050] Figure 19 This is a flowchart illustrating a target block prediction method using a bitstream according to an embodiment.

[0051] Figure 20 This shows the template configuration in the example inter-frame prediction.

[0052] Figure 21 This shows the template configuration in intra-frame prediction based on the example.

[0053] Figure 22 This shows the template configuration in the example bilateral matching.

[0054] Figure 23 This shows the first configuration of the template in the template matching example.

[0055] Figure 24 This shows the second configuration of the template in the template matching example.

[0056] Figure 25 This shows the first configuration of a template with alternating empty lines in the template matching example.

[0057] Figure 26 This shows a second configuration of a template with alternating empty lines in the template matching example.

[0058] Figure 27 This illustrates a second configuration of a template with alternating empty pixels in template matching based on the example.

[0059] Figure 28 This illustrates a second configuration of a template with alternating empty pixels in template matching based on the example.

[0060] Figure 29 This shows the template configuration in the example bilateral matching.

[0061] Figure 30 This shows the configuration of a template with alternating empty lines in a bilateral match based on the example.

[0062] Figure 31 This shows the configuration of a template with alternating empty pixels in a bilateral matching example.

[0063] Figure 32 This illustrates the template configuration when performing intra-frame prediction in inter-frame images, based on an example.

[0064] Figure 33 This shows the template configuration when performing inter-frame prediction on inter-frame images, based on an example.

[0065] Figure 34 This is a flowchart of the prediction method according to an embodiment.

[0066] Figure 35 This illustrates downsampling based on a deep neural network, as shown in the example.

[0067] Figure 36 This shows the range of sample values ​​divided into N groups based on the example.

[0068] Figure 37 The range of sample values ​​divided into two groups is shown based on the example.

[0069] Figure 38 The model is shown as a new group derived from the example for the average values ​​of the sample points.

[0070] Figure 39 This shows a model derived from the average values ​​of sample points for multiple groups, based on the example.

[0071] Figure 40The example shows the luminance blocks and the corresponding luminance blocks to the chrominance blocks.

[0072] Figure 41 Showing a chroma block based on the example.

[0073] Figure 42 The example shows a chromaticity prediction block.

[0074] Figure 43 The example shows the luminance block and the luminance block corresponding to the chrominance block.

[0075] Figure 44 Showing a chroma block based on the example.

[0076] Figure 45 The example shows four chromaticity prediction blocks.

[0077] Figure 46 The diagram shows the regions in the luminance block according to the example, as well as the regions in the luminance block that directly correspond to the chrominance block.

[0078] Figure 47 The area shown is within the chroma block according to the example.

[0079] Figure 48 This shows the area in the luminance block that directly corresponds to the chroma block, based on the example.

[0080] Figure 49 The area shown is within the chroma block according to the example.

[0081] Figure 50 The example shows the chromaticity prediction block and chromaticity template.

[0082] Figure 51 This is a flowchart based on the example chromaticity prediction method.

[0083] Figure 52 The table of chromaticity prediction modes based on the model is shown in the example.

[0084] Figure 53 The example shows how to reconstruct a prediction model table using the lowest error cost.

[0085] Figure 54 The example shows a table of predicted patterns reconstructed using an error cost below or equal to the average.

[0086] Figure 55 The gradient detection filter or gradient detection pattern is shown according to the example.

[0087] Figure 56 The example shows a gradient block derived using a brightness block and four gradient detection filters.

[0088] Figure 57 The example shows regions in the brightness block and regions in the gradient block derived from the four gradient detection filters.

[0089] Figure 58 The diagram shows a region within a gradient block derived using four gradient detection filters, based on the example.

[0090] Figure 59 The area shown is within the chroma block according to the example.

[0091] Figure 60 The example shows the chromaticity prediction template and chromaticity template derived using four gradient-based linear models.

[0092] Figure 61 This is a flowchart based on the example chromaticity prediction method.

[0093] Figure 62 The table of chromaticity prediction modes based on the model is shown in the example.

[0094] Figure 63 The example shows how to reconstruct a prediction model table using the lowest error cost.

[0095] Figure 64 The table shows a low-cost reconstruction prediction model based on the example, using errors below or equal to the average.

[0096] Figure 65 The example shows the form of luminance samples and applied filters corresponding to chrominance samples.

[0097] Figure 66 The chromaticity sample points are shown based on the example.

[0098] Figure 67 This shows the filter shape and the location of the filter coefficients based on the example.

[0099] Figure 68 The shape of the filter is shown according to the example.

[0100] Figures 69a to 69r The shapes of various filters are shown according to the examples.

[0101] Figure 70 The filter bank 1 is shown according to the example.

[0102] Figure 71 Filter bank 2 is shown according to the example.

[0103] Figure 72 The filter bank 1 is shown according to the example.

[0104] Figure 73 Filter bank 2 is shown according to the example.

[0105] Figure 74 Filter bank 3 is shown according to the example.

[0106] Figure 75 Filter bank 4 is shown according to the example.

[0107] Figure 76 The example illustrates how to determine the filter based on the direction of the intra-frame prediction mode.

[0108] Figure 77 The example shows the downsampled luminance sample and surrounding sample corresponding to the chrominance sample.

[0109] Figure 78 The example shows the form of luminance samples and applied filters corresponding to chrominance samples.

[0110] Figure 79 The chromaticity sample points are shown based on the example.

[0111] Figure 80 This shows the filter shape and the location of the filter coefficients based on the example.

[0112] Figure 81 The image shows the positions of the downsampled luminance samples and the corresponding chrominance samples, based on the example.

[0113] Figure 82 This shows another location of the downsampled luminance sample and the location of the corresponding chrominance sample, based on the example.

[0114] Figure 83 The example shows the downsampling filter F1().

[0115] Figure 84 The example shows the downsampling filter F2().

[0116] Figure 85 The example shows the downsampling filter F3().

[0117] Figure 86 The example shows the downsampling filter F4().

[0118] Figure 87 The example shows the downsampling filter F5().

[0119] Figure 88 The model shown is based on the example, and multiple pieces of information are used for the model.

[0120] Figure 89 This is the first graph showing the values ​​used to set the representative values ​​according to the example.

[0121] Figure 90This is a second graph showing the values ​​used to set representative values ​​according to the example.

[0122] Figure 91 This is a third graph showing the values ​​used to set representative values ​​based on the example.

[0123] Figure 92 The example illustrates how the number of reference lines and reference points are determined based on the partition shape of the block.

[0124] Figure 93 Shows the current block and its neighboring blocks based on the example.

[0125] Figure 94 Shows the current block and the reference area for the current block, based on the example.

[0126] Figure 95 The example illustrates how the number of reference lines and reference samples are determined based on the intra-frame prediction mode.

[0127] Figure 96 This illustrates how the reference area can be modified to include areas detected by the matching method, based on an example.

[0128] Figure 97 The example demonstrates adjusting the reference area based on the location indicated by the region detected by the matching method.

[0129] Figure 98 This shows the configuration of the reference area based on the example.

[0130] Figure 99 This shows a reference block for configuring the reference area, based on an example.

[0131] Figure 100 This shows a combination of reference areas based on the example.

[0132] Figure 101 The example shows the location of non-adjacent blocks used to determine the reference area.

[0133] Figure 102 The size and shape of the actual encoded / decoded block are shown in the example at the location of a non-adjacent block.

[0134] Figure 103 The locations of non-adjacent blocks are shown based on the example.

[0135] Figure 104 This shows the range of reference blocks to be selected based on the example.

[0136] Figure 105 The example shows the filtering of the predicted samples.

[0137] Figure 106This illustrates how to generate prediction blocks using a blending approach, as shown in the example.

[0138] Figure 107 The example shows filtering for the prediction block.

[0139] Figure 108 The example demonstrates how to generate a prediction block by utilizing sample points predicted by three models.

[0140] Figure 109 The example shows filtering for predicted samples and chroma blocks.

[0141] Figure 110 This shows the construction and classification of the candidate list based on the example.

[0142] Figure 111 This is a flowchart of the cross-component prediction merging method based on the example.

[0143] Figure 112 This shows the positions of the current block's neighboring and non-neighboring blocks, based on the example.

[0144] Figure 113 Shows other locations of the current block's neighboring and non-neighboring blocks, based on the example.

[0145] Figure 114 This shows the location of neighboring blocks within the current frame, based on the example.

[0146] Figure 115 This shows the positions of neighboring blocks within the current screen and reference screen, based on the example.

[0147] Figure 116 Show other locations of neighboring blocks within the current frame, based on the example.

[0148] Figure 117 Show other locations of neighboring blocks within the current screen and reference screen, based on the example.

[0149] Figure 118 The example shows the neighboring blocks at the location where the block vector is moved.

[0150] Figure 119 This shows the sorting of candidates based on the candidate list in the example.

[0151] Figure 120 The candidate list is shown, sorted by template matching cost based on the example.

[0152] Figure 121 The candidate list is shown, ranked by chromaticity prediction cost based on the example.

[0153] Figure 122The example demonstrates how prediction blocks are generated by using a weighted sum of prediction blocks generated from multiple models.

[0154] Figure 123 The example demonstrates how to generate a predicted signal by using signals generated from different cross-component prediction models for the sample points.

[0155] Figure 124 The U-predicted signal is shown as corrected using a color component prediction model based on the example, which uses the luminance signal and the chrominance signal.

[0156] Figure 125 The example shows the Y prediction signal corrected using a color component prediction model between the luminance and chrominance signals.

[0157] Figure 126 The U-predicted signal is shown as a result of a color component prediction model based on the example, using the chromaticity signals.

[0158] Figure 127 The V-predicted signal is shown as a result of a color component prediction model based on the example, using the chromaticity signals.

[0159] Figure 128 The gradient pattern is shown based on the example.

[0160] Figure 129 The example shows chromaticity samples used to derive a linear model.

[0161] Figure 130 The example shows the brightness samples used to derive the linear model.

[0162] Figure 131 The example shows the non-downsampled brightness sample points.

[0163] Figure 132 The filter for samples in CCLM is shown in the example.

[0164] Figure 133 This shows a template for a block based on the example.

[0165] Figure 134 The various downsampling filters used in the cross-component model are shown in the example.

[0166] Figure 135 This shows the location of the chromaticity sample points according to the example.

[0167] Figure 136 The spatial portion of the convolution filter is shown according to the example.

[0168] Figure 137 The reference area for BVG-CCCM is shown according to the example.

[0169] Figure 138 The example shows the location used to derive the block vector from the co-position brightness block.

[0170] Figure 139 The location of the time candidate is shown based on the example.

[0171] Figure 140 The position of the time candidate for shifting is shown according to the example.

[0172] Figure 141 This shows the location used to select the nearest motion vector.

[0173] Figure 142 This shows a time candidate for shift derived based on the current motion vector, as illustrated in the example.

[0174] Figure 143 The decoding method is shown based on the example.

[0175] Figure 144 The example shows the luminance samples related to the chromaticity samples.

[0176] Figure 145 This shows the template adjacent to the current chromaticity CU, based on the example. Detailed Implementation

[0177] This invention can be modified in various ways and can have various embodiments, which will be described in detail below with reference to the accompanying drawings. However, it should be understood that these embodiments are not intended to limit the invention to the specific forms disclosed, and they include all variations, equivalents, or modifications included within the spirit and scope of the invention.

[0178] The following exemplary embodiments will be described in detail with reference to the accompanying drawings, which illustrate specific embodiments. These embodiments are described to enable those skilled in the art to readily implement them. It should be noted that the various embodiments differ from one another but are not necessarily mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented as other embodiments without departing from the spirit and scope of other embodiments associated with one embodiment. Furthermore, it should be understood that the position or arrangement of various components in each disclosed embodiment can be changed without departing from the spirit and scope of the embodiments. Therefore, the appended detailed description is not intended to limit the scope of this disclosure, and the scope of the exemplary embodiments is defined only by the appended claims and their equivalents (provided they are properly described).

[0179] In the accompanying drawings, similar reference numerals are used to designate the same or similar functions in various respects. The shape, size, etc., of the components in the drawings may be exaggerated for clarity of description.

[0180] Terms such as “first” and “second” may be used to describe various components, but components are not limited by these terms. These terms are used only to distinguish one component from another. For example, without departing from the scope of this specification, a first component may be referred to as a second component. Similarly, a second component may be referred to as a first component. The term “and / or” may include a combination of multiple related descriptive terms or any one of multiple related descriptive terms.

[0181] It will be understood that when a component is referred to as "connected" or "combined" to another component, the two components may be directly connected or combined with each other, or there may be an intermediate component between the two components. On the other hand, it will be understood that when a component is referred to as "directly connected or combined," there is no intermediate component between the two components.

[0182] The components described in the embodiments are shown independently to indicate different functional features, but this does not mean that each component is formed by a single piece of hardware or software. That is, for ease of description, multiple components are arranged and included separately. For example, at least two of the multiple components may be integrated into a single component. Conversely, a component may be divided into multiple components. Embodiments in which multiple components are integrated or some components are separated are included within the scope of this specification, provided they do not depart from its spirit.

[0183] The terminology used in the embodiments is for describing particular embodiments only and is not intended to limit the invention. Singular expressions include plural expressions unless the context specifically indicates the contrary. In the embodiments, it should be understood that terms such as “comprising” or “having” are intended only to indicate the presence of features, numbers, steps, operations, components, parts, or combinations thereof, and are not intended to exclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof. That is, in the embodiments, the expression describing a component as “comprising” a specific component means that additional components may be included within the practice or technical spirit of the invention, but do not exclude the presence of components other than the specific component stated therein.

[0184] In embodiments, the term "at least one" may mean one of one or more quantities (such as 1, 2, 3, and 4). In embodiments, the term "a plurality of" may mean one of two or more quantities (such as 2, 3, and 4).

[0185] Some components of the embodiments are not essential components for performing the necessary functions, but may be optional components used only to improve performance. Embodiments may be implemented using only the essential components necessary to achieve the essence of the embodiments. For example, a structure that includes only the essential components (excluding optional components used only to improve performance) is also included within the scope of the embodiments.

[0186] The embodiments will now be described in detail with reference to the accompanying drawings, enabling those skilled in the art to readily implement the embodiments. In the following description of the embodiments, detailed descriptions of well-known functions or configurations that are considered to obscure key points of this specification will be omitted. Furthermore, the same reference numerals are used throughout the drawings to designate the same components, and repeated descriptions of the same components will be omitted.

[0187] In the following text, "image" may refer to a single frame that constitutes a video, or it may refer to the video itself. For example, "encoding and / or decoding of an image" may mean "encoding and / or decoding of a video," and may also mean "encoding and / or decoding of any one of the multiple images that constitute a video."

[0188] In the following text, the terms “video” and “moving footage” may be used to have the same meaning and may be used interchangeably.

[0189] In the following text, the target image can be an encoded target image that is the target to be encoded and / or a decoded target image that is the target to be decoded. Furthermore, the target image can be an input image input to an encoding device or an input image input to a decoding device. Also, the target image can be the current image, i.e., the target currently to be encoded and / or decoded. For example, the terms "target image" and "current image" can be used to have the same meaning and can be used interchangeably.

[0190] In the following text, the terms “image,” “picture,” “frame,” and “screen” may be used to have the same meaning and may be used interchangeably with each other.

[0191] In the following text, a target block can be an encoding target block (i.e., the target to be encoded) and / or a decoding target block (i.e., the target to be decoded). Furthermore, a target block can be the current block, i.e., the target currently to be encoded and / or decoded. Here, the terms "target block" and "current block" can be used to have the same meaning and are interchangeable. A current block can represent an encoding target block that is the encoding target during encoding and / or a decoding target block that is the decoding target during decoding. Furthermore, a current block can be at least one of an encoding block, a prediction block, a residual block, and a transform block.

[0192] In the following text, the terms “block” and “unit” may be used to have the same meaning and may be used interchangeably. Alternatively, “block” may refer to a specific unit.

[0193] In the following text, the terms “region” and “fragment” are used interchangeably.

[0194] In the following embodiments, specific information, data, flags, indexes, elements, and attributes may have their own values. The value "0" corresponding to each of the information, data, flags, indexes, elements, and attributes may indicate false, logical false, or a first predefined value. In other words, the values ​​"0," false, logical false, and the first predefined value are interchangeable. The value "1" corresponding to each of the information, data, flags, indexes, elements, and attributes may indicate true, logical true, or a second predefined value. In other words, the values ​​"1," true, logical true, and the second predefined value are interchangeable.

[0195] When variables such as i or j are used to indicate rows, columns, or indices, the value i can be an integer 0 or a greater than 0, or an integer 1 or a greater than 1. In other words, in the embodiments, each of the rows, columns, and indices can be counted starting from 0 or starting from 1.

[0196] In the embodiments, the term "one or more" or the term "at least one" may mean the term "multiple". The terms "one or more" or the term "at least one" may be used interchangeably with "multiple".

[0197] The terminology used in the embodiments will be described below.

[0198] Encoder: An encoder represents a device used to perform encoding. In other words, an encoder can represent an encoding device.

[0199] Decoder: A decoder refers to a device used to perform decoding. In other words, a decoder can represent a decoding device.

[0200] Unit: A unit can represent a component of image encoding and decoding. The terms "unit" and "block" can be used interchangeably and have the same meaning.

[0201] – A cell can be an M×N sample array. Each of M and N can be a positive integer. Cells can typically represent sample arrays in two-dimensional form.

[0202] In the process of image encoding and decoding, a "unit" can be a region generated by partitioning an image. In other words, a "unit" can be a specified region within an image. A single image can be partitioned into multiple units. Alternatively, an image can be partitioned into sub-parts, and a unit can represent each sub-part created when encoding or decoding is performed on the partitioned sub-parts.

[0203] – During the encoding and decoding of an image, predefined processing can be performed on each unit according to its type.

[0204] Based on function, unit types can be classified as macrounits, coding units (CUs), prediction units (PUs), residual units, transform units (TUs), etc. Alternatively, based on function, units can represent blocks, macroblocks, coding tree units, coding tree blocks, coding units, coding blocks, prediction units, prediction blocks, residual units, residual blocks, transform units, transform blocks, etc. For example, the target unit that serves as the object of encoding and / or decoding can be at least one of CUs, PUs, residual units, and TUs.

[0205] – The term “unit” can refer to a block of luma components, a block of chroma components corresponding to the luma components, and information about the syntax elements for each block, such that the unit is specified to be distinct from the block.

[0206] The size and shape of the unit can be implemented differently. In addition, the unit can have any of a variety of sizes and shapes. Specifically, the shape of the unit can include not only squares, but also geometric shapes that can be represented in two dimensions (2D) (such as rectangles, trapezoids, triangles and pentagons).

[0207] In addition, cell information may include one or more of the following: cell type, cell size, cell depth, cell encoding order, and cell decoding order. For example, the cell type may indicate one of CU, PU, ​​residual cell, and TU.

[0208] – A cell can be divided into sub-cells, each sub-cell having a smaller size than the related cell.

[0209] Depth: Depth represents the degree to which a cell is partitioned. Furthermore, cell depth indicates the level at which a cell exists when represented by a tree structure.

[0210] – Cell partitioning information may include depth, which indicates the depth of the cell. Depth may indicate the number of times the cell is partitioned and / or the degree to which the cell is partitioned.

[0211] In a tree structure, the root node can be considered to have the smallest depth and the leaf nodes have the largest depth. The root node can be the highest (top) node. The leaf nodes can be the lowest nodes.

[0212] A single unit can be hierarchically partitioned into multiple sub-units, with each unit possessing depth information based on a tree structure. In other words, a unit and the sub-units generated by partitioning that unit can correspond to a node and its child nodes, respectively. Each partitioned sub-unit can have a unit depth. Since depth indicates the number of times a unit is partitioned and / or the degree to which a unit is partitioned, the partitioning information of a sub-unit can include information about the size of that sub-unit.

[0213] In a tree structure, the top node can correspond to the initial node before partitioning. The top node can be called the "root node". Furthermore, the root node can have a minimum depth value. Here, the depth of the top node can be level "0".

[0214] A node with a depth of level "1" can represent a cell generated when the initial cell is partitioned once. A node with a depth of level "2" can represent a cell generated when the initial cell is partitioned twice.

[0215] – A leaf node of depth “n” can represent a cell generated when the initial cell is partitioned n times.

[0216] – A leaf node can be a bottom node that cannot be further partitioned. The depth of a leaf node can be the maximum level. For example, a predefined value for the maximum level could be 3.

[0217] –QT depth can represent the depth for a four-partition drive. BT depth can represent the depth for a two-partition drive. TT depth can represent the depth for a three-partition drive.

[0218] – Sample: A sample can be the basic unit that makes up a block. It can be 0 to 2 based on the bit depth (Bd). Bd The value of -1 is used to represent the sample point.

[0219] – A sample point can be a pixel or a pixel value.

[0220] – In the following text, the terms “pixel” and “sample” may be used to have the same meaning and may be used interchangeably.

[0221] Code Tree Unit (CTU): A CTU can consist of a single luma component (Y) code tree block and two chroma component (i.e., Cb, Cr) code tree blocks associated with the luma component code tree block. Furthermore, a CTU can represent information including the aforementioned blocks and the syntax elements used for each block.

[0222] – Each coding tree unit (CTU) can be partitioned using one or more partitioning methods (such as quadtree (QT), binary tree (BT), and ternary tree (TT)) to configure sub-units, such as coding units, prediction units, and transform units. A quadtree can represent a quaternion tree. Additionally, one or more partitioning methods can be used to partition each coding tree unit using multi-type tree (MTT).

[0223] – “CTU” can be used as a term to specify a pixel block as a processing unit in the process of image decoding and encoding (such as in the case of partitioning an input image).

[0224] Coding Tree Block (CTB): "CTB" can be used as a term to specify any one of the Y coding tree block, Cb coding tree block, and Cr coding tree block.

[0225] Neighboring Blocks: Neighboring blocks (or adjacent blocks) can represent blocks that are adjacent to the target block. Neighboring blocks can also represent neighboring blocks that are being reconstructed.

[0226] In the following text, the terms “nearby block” and “adjacent block” may be used to have the same meaning and may be used interchangeably.

[0227] Neighboring blocks can represent reconstructed neighboring blocks.

[0228] Spatial neighbor block: A spatial neighbor block can be a block that is spatially adjacent to the target block. Neighbor blocks can include spatial neighbor blocks.

[0229] – Target blocks and spatially adjacent blocks can be included in the target frame.

[0230] – A spatially adjacent block can represent a block whose boundary is in contact with the target block or a block located within a predetermined distance from the target block.

[0231] – A spatial neighbor block can represent a block that is adjacent to the vertex of the target block. Here, a block that is adjacent to the vertex of the target block can be a block that is vertically adjacent to a neighbor block that is horizontally adjacent to the target block, or a block that is horizontally adjacent to a neighbor block that is vertically adjacent to the target block.

[0232] Temporally neighboring blocks: Temporally neighboring blocks can be blocks that are temporally adjacent to the target block. Neighboring blocks can include temporally neighboring blocks.

[0233] –Time-proximity blocks can include col blocks.

[0234] The col block can be a block in a previously reconstructed co-location frame (col frame). The position of the col block in the col frame can correspond to the position of the target block in the target frame. Alternatively, the position of the col block in the col frame can be equal to the position of the target block in the target frame. The col frame can be a frame included in the list of reference frames.

[0235] - A temporally neighboring block can be a block that is spatially adjacent to the target block in time.

[0236] Prediction mode: Prediction mode can be information indicating the mode used for intra-frame prediction or the mode used for inter-frame prediction.

[0237] Prediction Unit: A prediction unit can be a basic unit used for prediction (such as inter-frame prediction, intra-frame prediction, inter-frame compensation, intra-frame compensation, and motion compensation).

[0238] A single prediction unit can be divided into multiple partitions or sub-prediction units with smaller sizes. These partitions can also be the basic units used in performing prediction or compensation. Partitions generated by dividing the prediction unit can also be prediction units.

[0239] Prediction cell partitioning: Prediction cell partitioning can be the shape in which prediction cells are divided.

[0240] Reconstructed neighboring cells: The reconstructed neighboring cells can be cells that are adjacent to the target cell and have already been decoded and reconstructed.

[0241] - The reconstructed neighboring units can be those that are spatially adjacent to the target unit or temporally adjacent to the target unit.

[0242] – The reconstructed spatial neighboring unit can be a unit included in the target image that has already been reconstructed through encoding and / or decoding.

[0243] – The reconstructed temporal neighbor unit can be a unit included in the reference image that has already been reconstructed through encoding and / or decoding. The position of the reconstructed temporal neighbor unit in the reference image can be the same as the position of the target unit in the target image, or it can correspond to the position of the target unit in the target image. Furthermore, the reconstructed temporal neighbor unit can be a block adjacent to a corresponding block in the reference image. Here, the position of the corresponding block in the reference image can correspond to the position of the target block in the target image. The fact that the positions of the blocks correspond to each other can indicate that the positions of the blocks are the same, that one block is included in another block, or that one block occupies a specific position in another block.

[0244] Sub-screen: A screen can be divided into one or more sub-screens. A sub-screen can consist of one or more parallel block rows and one or more parallel block columns.

[0245] – A sub-screen can be an area within a screen that has a square or rectangular shape (i.e., a non-square rectangle). Furthermore, a sub-screen can include one or more CTUs.

[0246] – A sub-picture can be a rectangular area of ​​one or more strips in the picture.

[0247] A sub-picture may include one or more parallel blocks, one or more bricks, and / or one or more stripes.

[0248] Parallel blocks: Parallel blocks can be areas in the image that have a square or rectangular shape (i.e., a non-square rectangle).

[0249] – A parallel block may include one or more CTUs.

[0250] – Parallel blocks can be partitioned into one or more blocks.

[0251] Blocking: Blocks can represent one or more CTU lines in a parallel block.

[0252] – A parallel block can be partitioned into one or more blocks. Each block may contain one or more CTU rows.

[0253] – Parallel blocks that are not partitioned into two parts can also be represented as blocks.

[0254] Strip: A strip may comprise one or more parallel blocks of a frame. Alternatively, a strip may comprise one or more sub-blocks of a parallel block.

[0255] A sub-picture may comprise one or more stripes that share a rectangular area that covers the entire picture. Therefore, each sub-picture boundary is always a stripe boundary, and each vertical sub-picture boundary is always a vertical parallel block boundary.

[0256] Parameter set: The parameter set corresponds to the header information in the internal structure of the bitstream.

[0257] – The parameter set may include at least one of the following: Video Parameter Set (VPS), Sequence Parameter Set (SPS), Picture Parameter Set (PPS), Adaptive Parameter Set (APS), Decoding Parameter Set (DPS).

[0258] Information sent via signals for each parameter set can be applied to the screen referencing the corresponding parameter set. For example, information in a VPS can be applied to the screen referencing the VPS. Information in an SPS can be applied to the screen referencing the SPS. Information in a PPS can be applied to the screen referencing the PPS.

[0259] Each parameter set can reference a higher-level parameter set. For example, PPS can reference SPS, and SPS can reference VPS.

[0260] Additionally, the parameter set may include parallel block groups, stripe header information, and parallel block header information. A parallel block group can be a group comprising multiple parallel blocks. Furthermore, the meaning of "parallel block group" can be the same as that of "strip".

[0261] Rate-distortion optimization: Coding devices can use rate-distortion optimization to provide high coding efficiency by utilizing a combination of the following: coding unit (CU) size, prediction mode, prediction unit (PU) size, motion information, and transform unit (TU) size.

[0262] – Rate distortion optimization schemes calculate the rate distortion cost of each combination to select the optimal combination. This can be achieved using equations such as "". To calculate the rate distortion cost, the combination that minimizes the rate distortion cost is typically chosen as the optimal combination under the rate distortion optimization scheme.

[0263] – D can represent distortion. D can be the average of the squares of the differences between the original transform coefficients and the reconstructed transform coefficients in the transform unit (i.e., mean square error).

[0264] – R can represent the rate, which can be expressed as a bit rate using relevant context information.

[0265] – R represents the Lagrange multiplier. It can include not only coding parameter information (such as prediction mode, motion information, and coding block flags), but also bits generated from encoding the transform coefficients.

[0266] The coding device can perform processes such as inter-frame prediction and / or intra-frame prediction, transform, quantization, entropy coding, inverse quantization (dequantization), and / or inverse transform to compute accurate D and R. These processes greatly increase the complexity of the coding device.

[0267] – Bitstream: A bitstream can represent a stream of bits that encode image information.

[0268] Parsing: Parsing can be the determination of the value of a syntax element by performing entropy decoding on a bitstream. Alternatively, the term "parsing" can refer to this entropy decoding itself.

[0269] Symbols: A symbol can be at least one of the syntax elements, encoding parameters, and transform coefficients of the encoding target unit and / or the decoding target unit. Furthermore, a symbol can be the target of entropy encoding or the result of entropy decoding.

[0270] Reference frame: The reference frame can be an image referenced by the cell to perform inter-frame prediction or motion compensation. Alternatively, the reference frame can be an image that includes reference cells referenced by the target cell to perform inter-frame prediction or motion compensation.

[0271] In the following text, the terms “reference screen” and “reference image” may be used to have the same meaning and may be used interchangeably.

[0272] Reference frame list: The reference frame list can be a list of one or more reference images that are used for inter-frame prediction or motion compensation.

[0273] – The types of reference screen lists can include combined list (LC), list 0 (L0), list 1 (L1), list 2 (L2), list 3 (L3), etc.

[0274] – For inter-frame prediction, one or more reference frame lists can be used.

[0275] Inter-frame prediction indicator: The inter-frame prediction indicator indicates the direction of inter-frame prediction for the target cell. Inter-frame prediction can be either unidirectional or bidirectional. Optionally, the inter-frame prediction indicator can represent the number of reference frames used to generate the prediction cell for the target cell. Optionally, the inter-frame prediction indicator can represent the number of prediction blocks used for inter-frame prediction or motion compensation of the target cell.

[0276] Prediction list utilization flags: Prediction list utilization flags indicate whether to use at least one reference screen from a specific reference screen list to generate prediction cells.

[0277] – Inter-frame prediction indicators can be derived using prediction list utilization flags. Conversely, prediction list utilization flags can be derived using inter-frame prediction indicators. For example, a prediction list utilization flag of "0" (as a first value) indicates that reference frames from the reference frame list are not used to generate prediction blocks for the target cell. A prediction list utilization flag of "1" (as a second value) indicates that the reference frame list is used to generate prediction cells for the target cell.

[0278] Reference screen index: The reference screen index can be an index that indicates a specific reference screen in the list of reference screens.

[0279] Screen Order Count (POC): The POC value of a screen indicates the order in which the corresponding screens are displayed.

[0280] Motion Vector (MV): A motion vector can be a 2D vector used for inter-frame prediction or motion compensation. A motion vector represents the offset between the target image and a reference image.

[0281] – For example, it can be in the form of (mv) x ,mv y MV is represented in the form of ) x It can indicate the horizontal component, mv y It can indicate the vertical component.

[0282] – Search Range: The search range can be a 2D region where a search for the MV is performed during inter-frame prediction. For example, the size of the search range can be M×N. M and N can both be positive integers.

[0283] Motion vector candidates: Motion vector candidates can be blocks that are used as prediction candidates when motion vectors are predicted, or motion vectors that are used as prediction candidates.

[0284] – Motion vector candidates can be included in the motion vector candidate list.

[0285] Motion vector candidate list: The motion vector candidate list can be a list using one or more motion vector candidate configurations.

[0286] Motion vector candidate index: The motion vector candidate index can be an indicator used to indicate motion vector candidates in the motion vector candidate list. Alternatively, the motion vector candidate index can be an index of motion vector predictors.

[0287] Motion information: Motion information may include at least one of the following: a list of reference frames, a reference image, motion vector candidates, a motion vector candidate index, a merge candidate and a merge index, as well as information on motion vectors, reference frame indexes and inter-frame prediction indicators.

[0288] Merge candidate list: The merge candidate list can be a list that uses one or more merge candidate configurations.

[0289] Merging Candidates: Merging candidates can be spatial merging candidates, temporal merging candidates, combined merging candidates, combined dual-prediction merging candidates, history-based candidates, average merging candidates based on the average of two candidates, zero merging candidates, etc. Merging candidates may include inter-frame prediction indicators and may include motion information such as prediction type information, reference frame index for each list, motion vectors, prediction list utilization flags, and inter-frame prediction indicators.

[0290] Merge index: A merge index can be an indicator used to indicate merge candidates in a merge candidate list.

[0291] - The merge index can indicate which reconstruction cell is used to derive the merge candidate among the reconstruction cells that are spatially adjacent to the target cell and temporally adjacent to the target cell.

[0292] – The merge index can indicate at least one of the multiple motion information candidates to be merged.

[0293] Transform Unit: A transform unit can be the basic unit for residual signal encoding and / or residual signal decoding (such as transform, inverse transform, quantization, dequantization, transform coefficient encoding, and transform coefficient decoding). A single transform unit can be partitioned into multiple sub-transform units with smaller sizes. Here, the transform may include one or more primary transforms and secondary transforms, and the inverse transform may include one or more primary inverse transforms and secondary inverse transforms.

[0294] Scaling: Scaling can represent the process of multiplying a factor by a transformation coefficient.

[0295] – As a result of scaling the levels of the transform coefficients, transform coefficients can be generated. Scaling can also be referred to as “inverse quantization”.

[0296] Quantization parameter (QP): The quantization parameter can be a value used to generate a transform coefficient level for the transform coefficients during quantization. Alternatively, the quantization parameter can also be a value used to generate transform coefficient values ​​by scaling the transform coefficient level during dequantization. Alternatively, the quantization parameter can be a value mapped to the quantization step size.

[0297] Delta quantization parameter: The Delta quantization parameter represents the difference between the quantization parameter of the target cell and the predicted quantization parameter.

[0298] Scan: A scan can refer to a method of arranging the coefficients in a cell, block, or matrix in order. For example, a method for arranging a 2D array in the form of a one-dimensional (1D) array can be called a "scan". Alternatively, a method for arranging a 1D array in the form of a 2D array can also be called a "scan" or "inverse scan".

[0299] Transform coefficients: Transform coefficients can be coefficient values ​​generated when the encoding device performs a transform. Alternatively, transform coefficients can be coefficient values ​​generated when the decoding device performs at least one of entropy decoding and dequantization.

[0300] – The level of quantization or the level of quantized transform coefficients generated by applying quantization to transform coefficients or residual signals can also be included in the meaning of the term “transform coefficients”.

[0301] Quantization level: The quantization level can be a value generated when the encoding device performs quantization on the transform coefficients or residual signal. Alternatively, the quantization level can be a value that serves as the target for dequantization when the decoding device performs dequantization.

[0302] – The transformation coefficient levels of quantization, as a result of transformation and quantization, can also be included in the meaning of the quantization levels.

[0303] Non-zero transform coefficients: Non-zero transform coefficients can be transform coefficients with values ​​other than 0, or transform coefficient levels with values ​​other than 0. Alternatively, non-zero transform coefficients can be transform coefficients with a value amplitude that is not zero, or transform coefficient levels with a value amplitude that is not zero.

[0304] Quantization matrix: A quantization matrix is ​​a matrix used during the quantization or dequantization process to improve the subjective or objective image quality of an image. A quantization matrix can also be referred to as a "scaling list".

[0305] Quantization matrix coefficients: Quantization matrix coefficients can be each element in the quantization matrix. Quantization matrix coefficients are also referred to as "matrix coefficients".

[0306] Default matrix: The default matrix can be a quantization matrix predefined by the encoding and decoding devices.

[0307] Non-default matrix: A non-default matrix can be a quantization matrix that is not predefined by the encoding and decoding devices. A non-default matrix can represent a quantization matrix sent by the user from the encoding device to the decoding device.

[0308] Most Probable Mode (MPM): MPM can represent an intra-prediction mode that is highly likely to be used for intra-prediction of the target block.

[0309] Encoding and decoding devices can determine one or more MPMs based on encoding parameters associated with the target block and attributes of entities associated with the target block.

[0310] Encoding and decoding devices can determine one or more MPMs based on the intra-prediction modes of a reference block. A reference block may include multiple reference blocks. These multiple reference blocks may include spatially adjacent blocks to the left of the target block and spatially adjacent blocks above the target block. In other words, one or more distinct MPMs can be determined based on which intra-prediction modes have been used for the reference blocks.

[0311] – One or more MPMs can be identified in the same way in both the encoding and decoding devices. That is, the encoding and decoding devices can share the same list of MPMs, which includes one or more MPMs.

[0312] MPM List: The MPM list can be a list that includes one or more MPMs. The number of one or more MPMs in the MPM list can be predefined.

[0313] MPM Indicator: The MPM indicator can indicate one or more MPMs in the MPM list that will be used for intra-prediction against the target block. For example, the MPM indicator can be an index used for the MPM list.

[0314] Since the MPM list is determined in the same way in both the encoding and decoding devices, it is not necessary to send the MPM list itself from the encoding device to the decoding device.

[0315] – The MPM indicator can be signaled from the encoding device to the decoding device. Because the MPM indicator is signaled, the decoding device can determine which MPM in the MPM list will be used for intra-frame prediction against the target block.

[0316] MPM Usage Indicator: The MPM usage indicator indicates whether an MPM usage mode will be used for prediction of the target block. The MPM usage mode can be determined using a list of MPMs to identify the MPMs that will be used for intra-frame prediction of the target block.

[0317] –MPM uses indicators that can be sent from the encoding device to the decoding device using signals.

[0318] Sending by signal: "Sending by signal" can indicate that information is sent from the encoding device to the decoding device. Alternatively, "sending by signal" can indicate that the information is included in a bitstream or recording medium by the encoding device. Information sent by signal by the encoding device can be used by the decoding device.

[0319] An encoding device generates encoded information by encoding the information to be transmitted as a signal. The encoded information is then sent from the encoding device to a decoding device. The decoding device obtains the information by decoding the transmitted encoded information. Here, the encoding can be entropy encoding, and the decoding can be entropy decoding.

[0320] Selective signal transmission: Information can be selectively transmitted using signals. Selective signal transmission for information can mean that an encoding device (under specific conditions) selectively includes information in a bitstream or recording medium. Selective signal transmission for information can also mean that a decoding device (under specific conditions) selectively extracts information from a bitstream.

[0321] The omission of "signal transmission": The use of "signal transmission" for information can be omitted. Regarding information, the omission of "signal transmission" for information may mean that the encoding device (under certain conditions) does not include the information in the bitstream or recording medium. The omission of "signal transmission" for information may also mean that the decoding device (under certain conditions) does not extract the information from the bitstream.

[0322] Statistical values: Variables, coding parameters, constants, etc., can have computable values. Statistical values ​​can be values ​​generated by performing calculations (operations) on the values ​​of a specified target. For example, a statistical value can indicate one or more of the following: the mean, weighted average, weighted sum, minimum, maximum, mode, median, and interpolation of the values ​​of a specific variable, a specific coding parameter, a specific constant, etc.

[0323] Figure 1 This is a block diagram illustrating the configuration of an embodiment of an encoding device to which the present disclosure is applied.

[0324] Encoding device 100 can be an encoder, a video encoding device, or an image encoding device. Video may include one or more images (frames). Encoding device 100 can sequentially encode one or more images of the video.

[0325] Reference Figure 1The encoding device 100 includes an inter-frame prediction unit 110, an intra-frame prediction unit 120, a switcher 115, a subtractor 125, a transform unit 130, a quantization unit 140, an entropy coding unit 150, an inverse quantization unit 160, an inverse transform unit 170, an adder 175, a filter unit 180, and a reference frame buffer 190.

[0326] The encoding device 100 can perform encoding on the target image using intra-frame mode and / or inter-frame mode. In other words, the prediction mode of the target block can be one of intra-frame mode and inter-frame mode.

[0327] In the following text, the terms "intra-frame mode", "intra-frame prediction mode", "in-frame mode" and "intra-frame prediction mode" may be used to have the same meaning and may be used interchangeably.

[0328] In the following text, the terms "inter-frame mode", "inter-frame prediction mode", "inter-picture mode" and "inter-picture prediction mode" may be used to have the same meaning and may be used interchangeably.

[0329] In the following text, the term "image" may refer to only a portion of an image, or it may refer to a block. Furthermore, the processing of an "image" may refer to the sequential processing of multiple blocks.

[0330] Furthermore, the encoding device 100 can generate a bitstream including encoded information by encoding the target image, and can output and store the generated bitstream. The generated bitstream can be stored in a computer-readable storage medium and can be streamed via wired and / or wireless transmission media.

[0331] When intra-frame mode is used as prediction mode, switcher 115 can switch to intra-frame mode. When inter-frame mode is used as prediction mode, switcher 115 can switch to inter-frame mode.

[0332] The encoding device 100 can generate a prediction block for the target block. Furthermore, after the prediction block has been generated, the encoding device 100 can use the residual between the target block and the prediction block to encode the residual block for the target block.

[0333] When the prediction mode is intra-frame mode, the intra-frame prediction unit 120 can use pixels from previously encoded / decoded neighboring blocks adjacent to the target block as reference samples. The intra-frame prediction unit 120 can use the reference samples to perform spatial prediction on the target block and can generate prediction samples for the target block via spatial prediction. The prediction samples can represent samples in the prediction block.

[0334] The inter-frame prediction unit 110 may include a motion prediction unit and a motion compensation unit.

[0335] When the prediction mode is inter-frame mode, the motion prediction unit can search for the region in the reference image that best matches the target block during motion prediction, and can derive motion vectors for both the target block and the found region based on the found region. Here, the motion prediction unit can use the search range as the target region for the search.

[0336] A reference image may be stored in a reference frame buffer 190. More specifically, when the encoding and / or decoding of a reference image has been processed, the encoded and / or decoded reference image may be stored in the reference frame buffer 190.

[0337] Since it stores the decoded screen, the reference screen buffer 190 can be a decoded screen buffer (DPB).

[0338] The motion compensation unit can generate a predicted block for the target block by performing motion compensation using motion vectors. Here, the motion vector can be a two-dimensional (2D) vector used for inter-frame prediction. Furthermore, the motion vector can indicate the offset between the target image and the reference image.

[0339] When the motion vector has values ​​other than integers, the motion prediction unit and motion compensation unit can generate prediction blocks by applying interpolation filters to a portion of the reference image. To perform inter-frame prediction or motion compensation, it can be determined which of the following modes—skip mode, merge mode, advanced motion vector prediction (AMVP) mode, and current frame reference mode—corresponds to the method for predicting and compensating for the motion of the PU included in the CU based on the CU, and inter-frame prediction or motion compensation can be performed according to that mode.

[0340] Subtractor 125 generates a residual block, which is the difference between the target block and the prediction block. The residual block can also be referred to as the "residual signal".

[0341] The residual signal can be the difference between the original signal and the predicted signal. Alternatively, the residual signal can be a signal generated by transforming or quantizing the difference between the original signal and the predicted signal, or by transforming and quantizing the difference. A residual block can be a residual signal for a block cell.

[0342] The transformation unit 130 can generate transformation coefficients by transforming the residual block, and can output the generated transformation coefficients. Here, the transformation coefficients can be coefficient values ​​generated by transforming the residual block.

[0343] Transformation unit 130 may use one of a number of predefined transformation methods when performing a transformation.

[0344] The predefined transformation methods may include Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Karhunen-Loeve Transform (KLT), etc.

[0345] The transformation method for transforming the residual block can be determined based on at least one of the coding parameters for the target block and / or neighboring blocks. For example, the transformation method can be determined based on at least one of the inter-frame prediction mode for the PU, the intra-frame prediction mode for the PU, the size of the TU, and the shape of the TU. Alternatively, transformation information indicating the transformation method can be transmitted from the encoding device 100 to the decoding device 200 by signaling.

[0346] When using the transform skip mode, the transform unit 130 can omit the operation of transforming the residual block.

[0347] By quantizing the transform coefficients, a quantized transform coefficient level or a quantized level can be generated. In the following examples, each of the quantized transform coefficient level and the quantized level may also be referred to as a "transform coefficient".

[0348] Quantization unit 140 can generate quantized transform coefficient levels (i.e., quantized levels or quantized coefficients) by quantizing the transform coefficients according to quantization parameters. Quantization unit 140 can output the generated quantized transform coefficient levels. In this case, quantization unit 140 can use a quantization matrix to quantize the transform coefficients.

[0349] Entropy coding unit 150 can generate a bitstream by performing probability distribution-based entropy coding based on values ​​calculated by quantization unit 140 and / or coding parameter values ​​calculated during the encoding process. Entropy coding unit 150 can output the generated bitstream.

[0350] The entropy coding unit 150 can perform entropy coding on information about the pixels of the image and information required to decode the image. For example, the information required to decode the image may include syntax elements, etc.

[0351] When applying entropy coding, fewer bits can be allocated to more frequently occurring symbols, and more bits can be allocated to less frequently occurring symbols. Because symbols are represented through this allocation, the size of the bit string used to encode the target symbol can be reduced. Therefore, entropy coding can improve the compression performance of video coding.

[0352] Furthermore, for entropy coding, the entropy coding unit 150 can use coding methods such as Exponential Golomb, Context Adaptive Variable Length Coding (CAVLC), or Context Adaptive Binary Arithmetic Coding (CABAC). For example, the entropy coding unit 150 can use a variable-length code / code (VLC) table to perform entropy coding. For example, the entropy coding unit 150 can derive a binarization method for the target symbol. Furthermore, the entropy coding unit 150 can derive a probabilistic model for the target symbol / bit. The entropy coding unit 150 can use the derived binarization method, probabilistic model, and context model to perform arithmetic coding.

[0353] The entropy coding unit 150 can transform the coefficients in 2D block form into 1D vector form through the transform coefficient scanning method, so as to encode the quantized transform coefficient levels.

[0354] Encoding parameters can be information required for encoding and / or decoding. Encoding parameters may include information encoded by encoding device 100 and sent from encoding device 100 to decoding device, and may also include information that can be derived during encoding or decoding. For example, information sent to decoding device may include syntax elements.

[0355] Encoding parameters can include not only information such as syntax elements (or flags or indexes) encoded by the encoding device and transmitted by the encoding device to the decoding device via signals, but also information derived during the encoding or decoding process. Furthermore, encoding parameters can include information required for encoding or decoding an image. For example, encoding parameters can include at least one value of the following, a combination of the following, or statistics of the following: cell / block size, cell / block shape / form, cell / block depth, cell / block partitioning information, cell / block partitioning structure, information indicating whether a cell / block is partitioned in a quadtree structure, information indicating whether a cell / block is partitioned in a binary tree structure, partitioning direction of the binary tree structure (horizontal or vertical), partitioning form of the binary tree structure (symmetric or asymmetric partitioning), information indicating whether a cell / block is partitioned in a ternary tree structure, partitioning direction of the ternary tree structure (horizontal or vertical), partitioning form of the ternary tree structure (symmetric or asymmetric partitioning), and partitioning direction of the ternary tree structure (horizontal or vertical). Information including: symmetric partitioning, whether the unit / block is partitioned in a multi-type tree structure, the combination and direction of partitions in the multi-type tree structure (horizontal or vertical, etc.), the partition form of the multi-type tree structure (symmetric or asymmetric partitioning, etc.), the partition tree in the multi-type tree form (binary tree or ternary tree), prediction type (intra-frame prediction or inter-frame prediction), intra-frame prediction mode / direction, intra-frame luma prediction mode / direction, intra-frame chroma prediction mode / direction, intra-frame partition information, inter-frame partition information, coded block partition flag, prediction block partition flag, transform block partition flag, reference sample filtering method, reference sample filter taps, reference sample filter coefficients, prediction block filtering method. Method, prediction block filter taps, prediction block filter coefficients, prediction block boundary filter method, prediction block boundary filter taps, prediction block boundary filter coefficients, inter-frame prediction mode, motion information, motion vector, motion vector difference, reference frame index, inter-frame prediction direction, inter-frame prediction indicator, prediction list utilization flag, reference frame list, reference image, POC, motion vector prediction factor, motion vector prediction index, motion vector prediction candidate, motion vector candidate list, information indicating whether the merge mode is used, merge index, merge candidate, merge candidate list, information indicating whether the skip mode is used, type of interpolation filter, interpolation filter taps, interpolation The filter parameters include: filter coefficients, motion vector magnitude, motion vector representation precision, transform type, transform size, information indicating whether a first transform is used, information indicating whether an additional (second) transform is used, first transform selection information (or first transform index), second transform selection information (or second transform index), information indicating the presence or absence of residual signals, code block pattern, code block flag, quantization parameters, residual quantization parameters, quantization matrix, information about the loop filter, information indicating whether the loop filter is applied, loop filter coefficients, loop filter taps, loop filter shape / form, and information indicating whether a deblocking filter is applied.Deblocking filter coefficients, deblocking filter taps, deblocking filter strength, deblocking filter shape / form, information indicating whether adaptive sample offset is applied, adaptive sample offset value, adaptive sample offset category, adaptive sample offset type, information indicating whether adaptive loop filter is applied, adaptive loop filter coefficients, adaptive loop filter taps, adaptive loop filter shape / form, binarization / debinarization method, context model, context model determination method, context model update method, information indicating whether normal mode is executed, information indicating whether bypass mode is executed, valid coefficient flag, last valid coefficient flag, coefficient group encoding flag, last valid coefficient position, information indicating whether the coefficient value is greater than 1, information indicating whether the coefficient value is greater than 2, information indicating whether the coefficient value is greater than 3, remaining coefficient value information, positive / negative sign information, reconstructed luminance sample, reconstructed chrominance sample, context binary bits. The following parameters are included: bypass binary bits, residual luminance samples, residual chrominance samples, transform coefficients, luminance transform coefficients, chrominance transform coefficients, quantization level, luminance quantization level, chrominance quantization level, transform coefficient level, transform coefficient level scanning method, size of the motion vector search area on the decoding device side, shape / form of the motion vector search area on the decoding device side, number of motion vector searches on the decoding device side, CTU size, minimum block size, maximum block size, maximum block depth, minimum block depth, image display / output order, stripe identification information, stripe type, stripe partition information, parallel block group identification information, parallel block group type, parallel block group partition information, parallel block identification information, parallel block type, parallel block partition information, screen type, bit depth, input sample bit depth, reconstructed sample bit depth, residual sample bit depth, transform coefficient bit depth, quantization level bit depth, information about the luminance signal, information about the chrominance signal, color space of the target block, and color space of the residual block. Furthermore, information related to the above encoding parameters may also be included in the encoding parameters. Information used to calculate and / or derive the above encoding parameters may also be included in the encoding parameters. Information calculated or derived using the above encoding parameters can also be included in the encoding parameters.

[0356] The first transformation selection information can indicate the first transformation to be applied to the target block.

[0357] The second transformation selection information can indicate the second transformation to be applied to the target block.

[0358] The residual signal can represent the difference between the original signal and the predicted signal. Alternatively, the residual signal can be a signal generated by transforming the difference between the original signal and the predicted signal. Alternatively, the residual signal can be a signal generated by transforming and quantizing the difference between the original signal and the predicted signal. A residual block can be a residual signal for a block.

[0359] Here, sending information via a signal can indicate that the encoding device 100 includes entropy-encoded information generated by performing entropy encoding on flags or indices in the bitstream, and can also indicate that the decoding device 200 obtains information by performing entropy decoding on the entropy-encoded information extracted from the bitstream. This information may include flags, indices, etc.

[0360] A signal can refer to information that will be transmitted using a signal. In the following text, information used for images and blocks may be referred to as a "signal". Furthermore, in the following text, the terms "information" and "signal" may be used to have the same meaning and may be used interchangeably. For example, a specific signal may be a signal representing a specific block. A raw signal may be a signal representing a target block. A prediction signal may be a signal representing a predicted block. A residual signal may be a signal representing a residual block.

[0361] The bitstream may include information based on a specific syntax. Encoding device 100 may generate a bitstream that includes information according to the specific syntax. Decoding device 200 may obtain information from the bitstream according to the specific syntax.

[0362] Since the encoding device 100 performs encoding via inter-frame prediction, the encoded target image can be used as a reference image for other images to be processed subsequently. Therefore, the encoding device 100 can reconstruct or decode the encoded target image and store the reconstructed or decoded image as a reference image in the reference frame buffer 190. For decoding, inverse quantization and inverse transform of the encoded target image can be performed.

[0363] The quantization levels can be dequantized by the dequantization unit 160 and inversely transformed by the inverse transform unit 170. The dequantization unit 160 can generate dequantized coefficients by performing an inverse transform on the quantization levels. The inverse transform unit 170 can generate coefficients that have undergone both dequantization and inverse transform by performing an inverse transform on the dequantized coefficients.

[0364] The coefficients that have undergone dequantization and inverse transform can be added to the prediction block by adder 175. Adding the coefficients that have undergone dequantization and inverse transform to the prediction block generates a reconstructed block. Here, the coefficients that have undergone dequantization and / or inverse transform can represent one or more coefficients that have undergone dequantization and inverse transform, and can also represent the reconstructed residual block. Here, the reconstructed block can represent either the recovered block or the decoded block.

[0365] The reconstructed blocks can be filtered by filter unit 180. Filter unit 180 can apply one or more filters, including deblocking filter, sample adaptive offset (SAO) filter, adaptive loop filter (ALF), and nonlocal filter (NLF), to the reconstructed samples, reconstructed blocks, or reconstructed images. Filter unit 180 may also be referred to as a "loop filter".

[0366] Deblocking filters eliminate block distortion that occurs at the boundaries between blocks in a reconstructed image. To determine whether to apply a deblocking filter, the number of columns or rows of pixels included in the block and on which the determination of whether to apply the deblocking filter to the target block is based can be determined.

[0367] When a deblocking filter is applied to a target block, the applied filter can vary depending on the required deblocking strength. In other words, among different filters, one that takes into account the strength of the deblocking filter can be applied to the target block. When a deblocking filter is applied to a target block, one or more filters, such as long-tap filters, strong filters, weak filters, and Gaussian filters, can be applied to the target block according to the required deblocking strength.

[0368] Furthermore, when performing vertical and horizontal filtering on the target block, horizontal and vertical filtering can be performed in parallel.

[0369] SAO can add an appropriate offset to the pixel value to compensate for coding errors. SAO can perform correction on the image to which deblocking is applied based on pixels, where the correction uses an offset of the difference between the original image and the image to which deblocking is applied. To perform offset correction on an image, methods can be used to divide the pixels included in the image into a specific number of regions, determine the region to be offset within the divided regions, and apply the offset to the determined region, or methods can be used to apply the offset taking into account the edge information of each pixel.

[0370] The ALF can perform filtering based on values ​​obtained by comparing the reconstructed image with the original image. After the pixels included in the image have been divided into a predetermined number of groups, the filter to be applied to each group can be determined, and filtering can be performed differently for each group. Information related to whether an adaptive loop filter is applied can be sent to each CU via a signal. This information can be sent via a signal for the luminance signal. The shape and filter coefficients of the ALF to be applied to each block can be different for each block. Alternatively, an ALF with a fixed form can be applied to the block regardless of its characteristics.

[0371] Nonlocal filters can perform filtering based on reconstructed blocks similar to the target block. Regions similar to the target block can be selected from the reconstructed image, and statistical properties of the selected similar regions can be used to perform filtering on the target block. Information regarding whether a nonlocal filter is applied can be sent to the coding unit (CU) via a signal. Furthermore, the shape and filter coefficients of the nonlocal filter applied to the block can vary depending on the block.

[0372] The reconstructed blocks or reconstructed image filtered by filter unit 180 can be stored as a reference frame in reference frame buffer 190. The reconstructed blocks filtered by filter unit 180 can be part of the reference frame. In other words, the reference frame can be a reconstructed frame composed of reconstructed blocks filtered by filter unit 180. The stored reference frame can then be used for inter-frame prediction or motion compensation.

[0373] Figure 2 This is a block diagram illustrating the configuration of an embodiment of the decoding device to which this disclosure is applied.

[0374] Decoding device 200 can be a decoder, video decoding device, or image decoding device.

[0375] Reference Figure 2 The decoding device 200 may include an entropy decoding unit 210, an inverse quantization unit 220, an inverse transform unit 230, an intra-frame prediction unit 240, an inter-frame prediction unit 250, a switcher 245, an adder 255, a filter unit 260, and a reference frame buffer 270.

[0376] Decoding device 200 can receive bit streams output from encoding device 100. Decoding device 200 can receive bit streams stored in computer-readable storage media and can also receive bit streams transmitted via wired / wireless transmission media.

[0377] The decoding device 200 can perform decoding on the bitstream in intra-frame mode and / or inter-frame mode. Furthermore, the decoding device 200 can generate a reconstructed image or a decoded image via decoding, and can output the reconstructed image or the decoded image.

[0378] For example, switcher 245 can be used to switch between intra-frame mode and inter-frame mode based on the prediction mode used for decoding. When the prediction mode used for decoding is intra-frame mode, switcher 245 can be operated to switch to intra-frame mode. When the prediction mode used for decoding is inter-frame mode, switcher 245 can be operated to switch to inter-frame mode.

[0379] The decoding device 200 can obtain the reconstructed residual block by decoding the input bitstream and can generate the prediction block. When the reconstructed residual block and the prediction block are obtained, the decoding device 200 can generate the reconstructed block, which is the target to be decoded, by adding the reconstructed residual block and the prediction block.

[0380] The entropy decoding unit 210 can generate symbols by performing entropy decoding on the bitstream based on the probability distribution of the bitstream. The generated symbols may include symbols in the form of quantized transform coefficient levels (i.e., quantized levels or quantized coefficients). Here, the entropy decoding method may be similar to the entropy coding method described above. That is, the entropy decoding method may be the inverse process of the entropy coding method described above.

[0381] The entropy decoding unit 210 can transform coefficients in one-dimensional (1D) vector form into 2D block shapes by a transform coefficient scanning method in order to decode the quantized transform coefficient levels.

[0382] For example, the coefficients of a block can be transformed into a 2D block shape by scanning the block coefficients using a top-right diagonal scan. Alternatively, which of the top-right diagonal scan, vertical scan, and horizontal scan will be used can be determined based on the size of the corresponding block and / or the intra-frame prediction mode.

[0383] The quantized coefficients can be dequantized by the dequantization unit 220. The dequantization unit 220 generates dequantized coefficients by performing dequantization on the quantized coefficients. Furthermore, the dequantized coefficients can be inversely transformed by the inverse transform unit 230. The inverse transform unit 230 generates a reconstructed residual block by performing an inverse transform on the dequantized coefficients. As a result of performing dequantization and inverse transform on the quantized coefficients, a reconstructed residual block can be generated. Here, when generating the reconstructed residual block, the dequantization unit 220 can apply the quantization matrix to the quantized coefficients.

[0384] When using intra-frame mode, intra-frame prediction unit 240 can generate a prediction block by performing spatial prediction on the target block, wherein the spatial prediction uses the pixel values ​​of previously decoded neighboring blocks adjacent to the target block.

[0385] The inter-frame prediction unit 250 may include a motion compensation unit. Alternatively, the inter-frame prediction unit 250 may be designated as a "motion compensation unit".

[0386] When using inter-frame mode, the motion compensation unit can generate a prediction block by performing motion compensation on the target block, wherein the motion compensation uses motion vectors and a reference image stored in the reference frame buffer 270.

[0387] The motion compensation unit can apply an interpolation filter to a portion of the reference image when the motion vector has values ​​other than integers, and can use the reference image with the interpolation filter applied to generate prediction blocks. To perform motion compensation, the motion compensation unit can determine, based on the CU, which of the following modes—skip mode, merge mode, advanced motion vector prediction (AMVP) mode, and current frame reference mode—corresponds to the motion compensation method used by the PU included in the CU, and can perform motion compensation according to the determined mode.

[0388] The reconstructed residual block and the predicted block can be added to each other by adder 255. Adder 255 generates a reconstructed block by adding the reconstructed residual block and the predicted block.

[0389] The reconstructed blocks can be filtered by filter unit 260. Filter unit 260 can apply at least one of a deblocking filter, a SAO filter, an ALF filter, and an NLF filter to the reconstructed blocks or the reconstructed image. The reconstructed image can be a picture that includes the reconstructed blocks.

[0390] The filter unit can output a reconstructed image.

[0391] The reconstructed image and / or reconstructed blocks filtered by filter unit 260 can be stored as a reference image in reference image buffer 270. The reconstructed blocks filtered by filter unit 260 can be part of the reference image. In other words, the reference image can be an image composed of reconstructed blocks filtered by filter unit 260. The stored reference image can then be used for inter-frame prediction or motion compensation.

[0392] Figure 3 It is a schematic diagram illustrating the partitioning structure of an image when it is encoded and decoded.

[0393] Figure 3 An example can be illustrated by showing a single cell divided into multiple sub-cells.

[0394] To effectively partition an image, coding units (CUs) can be used in encoding and decoding. The term "unit" can be used to collectively specify 1) a block comprising image samples and 2) a syntax element. For example, "partition of a unit" can mean "partition of a block corresponding to a unit".

[0395] A CU can be used as the basic unit for image encoding / decoding. A CU can be used as the unit to which one of the intra-frame and inter-frame modes is applied during image encoding / decoding. In other words, during image encoding / decoding, it can be determined which of the intra-frame and inter-frame modes will be applied to each CU.

[0396] Furthermore, the CU can be the basic unit for predicting, transforming, quantizing, inverse transforming, dequantizing, and encoding / decoding transform coefficients.

[0397] Reference Figure 3 Image 300 can be sequentially partitioned into units corresponding to the largest coding unit (LCU), and the partitioning structure can be determined for each LCU. Here, LCU can be used to have the same meaning as coding tree unit (CTU).

[0398] Partitioning a cell can represent partitioning the block corresponding to the cell. Block partitioning information may include depth information about the depth of the cell. The depth information may indicate the number of times the cell is partitioned and / or the degree to which the cell is partitioned. A single cell may be hierarchically partitioned into multiple sub-cells, and the single cell may have depth information based on a tree structure.

[0399] Each partitioned sub-unit can have depth information. The depth information can be information indicating the size of the CU. Depth information can be stored for each CU.

[0400] Each CU can have depth information. When a CU is partitioned, the depth of the CU generated from the partition can be increased by 1 from the depth of the partitioned CU.

[0401] The partitioning structure represents the distribution of coding units (CUs) in the LCU 310 used for effective encoding of an image. This distribution can be determined by whether a single CU will be partitioned into multiple CUs. The number of CUs generated by partitioning can be a positive integer of 2 or greater, including 2, 3, 4, 8, 16, etc.

[0402] Depending on the number of CUs generated through partitioning, the horizontal and vertical dimensions of each CU generated through partitioning can be smaller than the horizontal and vertical dimensions of the CU before partitioning. For example, the horizontal and vertical dimensions of each CU generated through partitioning can be half the horizontal and vertical dimensions of the CU before partitioning.

[0403] Each partitioned CU can be recursively partitioned into four CUs in the same manner. Through recursive partitioning, at least one of the horizontal and vertical dimensions of each partitioned CU can be reduced compared to at least one of the horizontal and vertical dimensions of the CU before partitioning.

[0404] The partitioning of a CU can be performed recursively until a predefined depth or predefined size is reached.

[0405] For example, the depth of a CU can range from 0 to 3. The size of a CU can range from 64×64 to 8×8, depending on its depth.

[0406] For example, the depth of LCU 310 can be 0, and the depth of the minimum coding unit (SCU) can be a predefined maximum depth. Here, as mentioned above, the LCU can be a CU with the maximum coding unit size, and the SCU can be a CU with the minimum coding unit size.

[0407] Partitioning can begin at LCU 310, and the depth of the CU can be increased by 1 whenever the horizontal and / or vertical dimensions of the CU are reduced by partitioning.

[0408] For example, for each depth, an unpartitioned CU can have a size of 2N×2N. Furthermore, when CUs are partitioned, a CU of size 2N×2N can be partitioned into four CUs, each with a size of N×N. The value of N is halved each time the depth increases by 1.

[0409] Reference Figure 3 An LCU with a depth of 0 can have 64×64 pixels or 64×64 blocks. 0 can be the minimum depth. An SCU with a depth of 3 can have 8×8 pixels or 8×8 blocks. 3 can be the maximum depth. Here, a CU with 64×64 blocks as an LCU can be represented by depth 0. A CU with 32×32 blocks can be represented by depth 1. A CU with 16×16 blocks can be represented by depth 2. A CU with 8×8 blocks as an SCU can be represented by depth 3.

[0410] Information about whether a corresponding CU is partitioned can be represented by the CU's partition information. Partition information can be 1 bit. All CUs except the SCU can include partition information. For example, the partition information value for an unpartitioned CU can be a first value. The partition information value for a partitioned CU can be a second value. When the partition information indicates whether a CU is partitioned, the first value can be "0" and the second value can be "1".

[0411] For example, when a single CU is partitioned into four CUs, the horizontal and vertical dimensions of each of the four CUs created through partitioning can be half the horizontal and vertical dimensions of the CU before partitioning. When a 32×32 CU is partitioned into four CUs, the size of each of the four partitioned CUs can be 16×16. When a single CU is partitioned into four CUs, the CU can be considered to have been partitioned using a quadtree structure. In other words, quadtree partitioning can be considered to have been applied to the CU.

[0412] For example, when a single CU is partitioned into two CUs, the horizontal or vertical dimension of each of the two resulting CUs can be half the horizontal or vertical dimension of the CU before partitioning. When a 32×32 CU is vertically partitioned into two CUs, the size of each of the two resulting CUs can be 16×32. When a 32×32 CU is horizontally partitioned into two CUs, the size of each of the two resulting CUs can be 32×16. When a single CU is partitioned into two CUs, the CU can be considered to have been partitioned using a binary tree structure. In other words, binary tree partitioning can be considered to have been applied to the CU.

[0413] For example, when a single CU is partitioned (or divided) into three CUs, the original CU before partitioning is partitioned such that its horizontal or vertical dimensions are divided in a 1:2:1 ratio, thus enabling the generation of three sub-CUs. For example, when a 16×32 CU is horizontally partitioned into three sub-CUs, the three sub-CUs generated by the partitioning can have dimensions of 16×8, 16×16, and 16×8 respectively from top to bottom. For example, when a 32×32 CU is vertically partitioned into three sub-CUs, the three sub-CUs generated by the partitioning can have dimensions of 8×32, 16×32, and 8×32 respectively from left to right. When a single CU is partitioned into three CUs, the CU can be considered to be partitioned in the form of a ternary tree. In other words, ternary tree partitioning can be considered to have been applied to the CU.

[0414] Quadtree partitioning and binary tree partitioning are both used... Figure 3 LCU 310.

[0415] In the encoding device 100, a 64×64 coding tree unit (CTU) can be partitioned into multiple smaller CUs using a recursive quadtree structure. A single CU can be partitioned into four CUs of the same size. Each CU can be recursively partitioned and can have a quadtree structure.

[0416] By using recursive partitioning of the CU, the optimal partitioning method that causes the minimum rate distortion cost can be selected.

[0417] Figure 3 The Coding Tree Unit (CTU) 320 in the example is a CTU in which quadtree partitioning, binary tree partitioning and ternary tree partitioning are all applied.

[0418] As described above, to partition a CTU, at least one of quadtree partitioning, binary tree partitioning, and ternary tree partitioning can be applied to the CTU. Partitioning can be applied based on specific priorities.

[0419] For example, quadtree partitioning can be preferentially applied to CTUs. CUs that cannot be further partitioned in quadtree form can correspond to the leaf nodes of a quadtree. CUs corresponding to the leaf nodes of a quadtree can be the root nodes of a binary tree and / or a ternary tree. That is, CUs corresponding to the leaf nodes of a quadtree can be partitioned in binary or ternary tree form, or may not be further partitioned. In this case, to prevent each CU generated by applying binary or ternary tree partitioning to the CUs corresponding to the leaf nodes of a quadtree from being quadtree partitioned again, the operations of block partitioning and / or signaling block partitioning information are effectively performed.

[0420] Four-partition information can be used to signal the partitions of a CU corresponding to each node of a quadtree. A four-partition message with a first value (e.g., "1") indicates that the corresponding CU is partitioned in quadtree form. A four-partition message with a second value (e.g., "0") indicates that the corresponding CU is not partitioned in quadtree form. The four-partition message can be a flag with a specific length (e.g., 1 bit).

[0421] There may be no priority between binary tree partitioning and ternary tree partitioning. That is, the CU corresponding to the leaf node of a quadtree can be partitioned in either binary or ternary tree form. Furthermore, CUs generated by binary or ternary tree partitioning can be further partitioned in either binary or ternary tree form, or they may not be further partitioned.

[0422] A partition executed when there is no priority between a binary tree partition and a ternary tree partition can be called a "multi-type tree partition". That is, the CU corresponding to a leaf node of a quadtree can be the root node of a multi-type tree. The partitioning of the CU corresponding to each node of the multi-type tree can be signaled using at least one of the following: information indicating whether the CU is partitioned according to the multi-type tree, partitioning direction information, and partitioning tree information. For the partitioning of the CU corresponding to each node of the multi-type tree, the information indicating whether the multi-type tree partitioning is executed, the partitioning direction information, and the partitioning tree information can be signaled sequentially.

[0423] For example, information indicating whether a CU is partitioned in a multi-type tree and having a first value (e.g., "1") can indicate that the corresponding CU is partitioned in a multi-type tree format. Information indicating whether a CU is partitioned in a multi-type tree and having a second value (e.g., "0") can indicate that the corresponding CU is not partitioned in a multi-type tree format.

[0424] When the CU corresponding to each node of the multi-type tree is partitioned in the form of a multi-type tree, the corresponding CU may further include partitioning direction information.

[0425] Partition direction information indicates the partitioning direction of a multi-type tree partition. Partition direction information with a first value (e.g., "1") indicates that the corresponding CU is partitioned in the vertical direction. Partition direction information with a second value (e.g., "0") indicates that the corresponding CU is partitioned in the horizontal direction.

[0426] When a CU corresponding to each node of a multi-type tree is partitioned in the form of a multi-type tree, the corresponding CU may further include partition tree information. The partition tree information can indicate the tree used for multi-type tree partitioning.

[0427] For example, partition tree information with a first value (e.g., "1") can indicate that the corresponding CU is partitioned in a binary tree format. Partition tree information with a second value (e.g., "0") can indicate that the corresponding CU is partitioned in a ternary tree format.

[0428] Here, each of the above-mentioned information indicating whether partitioning by multi-type tree is performed, partition tree information, and partition direction information can be a flag with a specific length (e.g., 1 bit).

[0429] Entropy encoding and / or entropy decoding can be performed on at least one of the above four partition information, information indicating whether partitioning according to multiple tree types has been executed, partition direction information, and partition tree information. To perform entropy encoding / entropy decoding of this information, information from neighboring CUs adjacent to the target CU can be used.

[0430] For example, it can be assumed that the partitioning patterns (i.e., partitioned / non-partitioned, partitioned tree, and / or partitioned direction) of the left and / or upper CUs are highly similar to the partitioning patterns of the target CU. Therefore, based on the information of neighboring CUs, contextual information for entropy encoding and / or entropy decoding of the information for the target CU can be derived. Here, the information of neighboring CUs may include at least one of the following: 1) four-partition information of neighboring CUs, 2) information indicating whether neighboring CUs are partitioned according to multiple types of trees, 3) partitioned direction information of neighboring CUs, and 4) partitioned tree information of neighboring CUs.

[0431] In another embodiment of binary tree partitioning and ternary tree partitioning, binary tree partitioning can be performed first. That is, binary tree partitioning can be applied first, and then the CUs corresponding to the leaf nodes of the binary tree can be set as the root nodes of the ternary tree. In this case, quadtree partitioning or binary tree partitioning may not be performed on the CUs corresponding to the nodes of the ternary tree.

[0432] A CU that is not further partitioned by quadtree partitioning, binary tree partitioning, and / or ternary tree partitioning can be a unit for encoding, prediction, and / or transformation. That is, a CU may not be further partitioned for use in prediction and / or transformation. Therefore, the partitioning structure used to partition CUs into prediction units (PUs) and / or transformation units (TUs), its partitioning information, etc., may not exist in the bitstream.

[0433] However, when the size of a CU (Computer Unit) used as a partitioning unit is larger than the size of the largest transform block, the CU can be recursively partitioned until the size of the CU becomes smaller than or equal to the size of the largest transform block. For example, when the size of the CU is 64×64 and the size of the largest transform block is 32×32, the CU can be partitioned into four 32×32 blocks to perform the transform. Similarly, when the size of the CU is 32×64 and the size of the largest transform block is 32×32, the CU can be partitioned into two 32×32 blocks.

[0434] In this case, it is not necessary to separately send a signal indicating whether the CU has been partitioned for transformation. Without signal transmission, partitioning of the CU can be determined by comparing its horizontal (and / or vertical) dimensions with the horizontal (and / or vertical) dimensions of the largest transform block. For example, when the horizontal dimension of the CU is greater than the horizontal dimension of the largest transform block, the CU can be vertically bisected. Furthermore, when the vertical dimension of the CU is greater than the vertical dimension of the largest transform block, the CU can be horizontally bisected.

[0435] Information regarding the maximum and / or minimum size of the CU and the maximum and / or minimum size of the transform block can be signaled or determined at a level higher than the CU level. For example, higher levels could be sequence level, frame level, parallel block level, parallel block group level, or stripe level. For example, the minimum size of the CU could be set to 4×4. For example, the maximum size of the transform block could be set to 64×64. For example, the maximum size of the transform block could be set to 4×4.

[0436] Information regarding the minimum size of the CU corresponding to the leaf node of the quadtree (i.e., the minimum size of the quadtree) and / or the maximum depth of the path from the root node to the leaf node of the multi-type tree (i.e., the maximum depth of the multi-type tree) can be signaled or determined at a level higher than the CU level. For example, higher levels could be sequence level, frame level, stripe level, parallel block group level, or parallel block level. Information regarding the minimum size of the quadtree and / or the maximum depth of the multi-type tree can be signaled or determined individually at each of the intra-strip and inter-strip levels.

[0437] Information regarding the difference between the size of the CTU and the maximum size of the transform block can be signaled or determined at a level higher than the CU level. For example, higher levels could be sequence level, frame level, stripe level, parallel block group level, or parallel block level. Information regarding the maximum size of the CU corresponding to each node of the binary tree (i.e., the maximum size of the binary tree) can be determined based on the size of the CTU and the aforementioned difference. The maximum size of the CU corresponding to each node of the ternary tree (i.e., the maximum size of the ternary tree) can have different values ​​depending on the stripe type. For example, the maximum size of the ternary tree in an intra-strip level could be 32×32. For example, the maximum size of the ternary tree in an inter-strip level could be 128×128. For example, the minimum size of the CU corresponding to each node of the binary tree (i.e., the minimum size of the binary tree) and / or the minimum size of the CU corresponding to each node of the ternary tree (i.e., the minimum size of the ternary tree) can be set to the minimum size of the CU.

[0438] In another example, the maximum size of a binary tree and / or the maximum size of a ternary tree can be signaled or determined at the stripe level. Furthermore, the minimum size of a binary tree and / or the minimum size of a ternary tree can be signaled or determined at the stripe level.

[0439] Based on the various block sizes and depths described above, the four partition information, information indicating whether partitioning by multiple tree types has been performed, partition tree information, and / or partition direction information may or may not exist in the bitstream.

[0440] For example, when the size of the CU is not greater than the minimum size of the quadtree, the CU may not include the four-partition information, and the four-partition information of the CU can be inferred as the second value.

[0441] For example, when the size (horizontal and vertical dimensions) of the CU corresponding to each node of the multi-type tree is greater than the maximum size (horizontal and vertical dimensions) of the binary tree and / or the maximum size (horizontal and vertical dimensions) of the ternary tree, the CU may not be partitioned in binary and / or ternary tree form. In this determination method, information indicating whether partitioning by multi-type tree is performed may not be sent by signal, but can be inferred as a second value.

[0442] Alternatively, when the size (horizontal and vertical dimensions) of the CU corresponding to each node of the multi-type tree is equal to the minimum size (horizontal and vertical dimensions) of the binary tree, or when the size (horizontal and vertical dimensions) of the CU is twice the minimum size (horizontal and vertical dimensions) of the ternary tree, the CU may not be partitioned in binary and / or ternary tree form. With this determination method, information indicating whether partitioning by multi-type tree is performed can be sent without signaling, but can be inferred as a second value. This is because when the CU is partitioned in binary and / or ternary tree form, it generates CUs smaller than the minimum size of the binary tree and / or the minimum size of the ternary tree.

[0443] Alternatively, binary or ternary partitioning can be limited based on the size of the virtual pipeline data unit (i.e., the size of the pipeline buffer). For example, binary or ternary partitioning can be limited when a CU is partitioned into sub-CUs that do not fit the size of the pipeline buffer. The size of the pipeline buffer can be equal to the maximum size of the transform block (e.g., 64×64).

[0444] For example, when the size of the pipeline buffer is 64×64, the following partitions can be restricted.

[0445] - A ternary tree partition for an N×M CU (where N and / or M are 128). - Horizontal binary tree partitioning for a 128×N CU (where N<=64) - Vertical binary tree partitioning for N×128 CUs (where N<=64) Alternatively, when the depth of the CU corresponding to each node of the multi-type tree is equal to the maximum depth of the multi-type tree, the CU may not be partitioned in binary and / or ternary tree form. With this determination method, information indicating whether partitioning by multi-type tree is performed can be sent without signaling, but can be inferred as a second value.

[0446] Alternatively, information indicating whether partitioning by the multi-type tree has been performed may be signaled only if at least one of vertical binary tree partitioning, horizontal binary tree partitioning, vertical ternary tree partitioning, and horizontal ternary tree partitioning is possible for each CU corresponding to each node of the multi-type tree. Otherwise, the CU may not be partitioned in binary and / or ternary tree form. In this determination method, the information indicating whether partitioning by the multi-type tree has been performed may not be signaled, but may be inferred as a second value.

[0447] Alternatively, for each CU corresponding to a node in a multi-type tree, partitioning direction information may be signaled only when both vertical binary tree partitioning and horizontal binary tree partitioning are feasible, or only when both vertical ternary tree partitioning and horizontal ternary tree partitioning are feasible. Otherwise, partitioning direction information may not be signaled, but may be inferred as a value indicating the direction in which the CU can be partitioned.

[0448] Alternatively, for each CU corresponding to a node of a multi-type tree, partition tree information may be signaled only when both vertical binary tree partitioning and vertical ternary tree partitioning are feasible, or only when both horizontal binary tree partitioning and horizontal ternary tree partitioning are feasible. Otherwise, partition tree information may not be signaled, but may be inferred as the value of the tree indicating the partitions that can be applied to the CU.

[0449] Figure 4 This is a diagram showing the form of prediction units that a coding unit can include.

[0450] Within the control units (CUs) partitioned from the control unit (LCU), CUs that are no longer partitioned can be divided into one or more prediction units (PUs). This partitioning is also known as "partitioning".

[0451] A PU (Program Unit) can be the basic unit used for prediction. A PU can be encoded and decoded in any of the following modes: skip mode, inter-frame mode, and intra-frame mode. A PU can be partitioned into various shapes according to each mode. For example, refer to the above... Figure 1 The target block described and the above references Figure 2 The target blocks described can all be PUs.

[0452] A CU may not be classified as a PU. When a CU is not classified as a PU, the dimensions of the CU and the PU can be equal.

[0453] In skip mode, partitioning may not be present in the CU. Skip mode also supports a 2N×2N mode 410 without partitioning, where the PU and CU have the same size.

[0454] In inter-frame mode, eight types of partition shapes can exist in the CU. For example, in inter-frame mode, 2N×2N mode 410, 2N×N mode 415, N×2N mode 420, N×N mode 425, 2N×nU mode 430, 2N×nD mode 435, nL×2N mode 440 and nR×2N mode 445 are supported.

[0455] In intra-frame mode, 2N×2N mode 410 and N×N mode 425 are supported.

[0456] In 2N×2N mode 410, a PU with a size of 2N×2N can be encoded. A PU with a size of 2N×2N can represent a PU with the same size as the CU. For example, a PU with a size of 2N×2N can have a size of 64×64, 32×32, 16×16, or 8×8.

[0457] In N×N mode 425, PUs with an N×N size can be encoded.

[0458] For example, in intra-frame prediction, when the PU size is 8×8, the PUs from four partitions can be encoded. The size of the PU from each partition can be 4×4.

[0459] When encoding a PU in intra-frame mode, the PU can be encoded using any of a number of intra-frame prediction modes. For example, HEVC technology provides 35 intra-frame prediction modes, and the PU can be encoded in any of these 35 intra-frame prediction modes.

[0460] The rate-distortion cost can be used to determine which of the 2N×2N modes 410 and N×N modes 425 will be used to encode the PU.

[0461] Encoding device 100 can perform encoding operations on a PU of size 2N×2N. Here, the encoding operation can be an operation of encoding the PU in each of a plurality of intra-prediction modes that can be used by encoding device 100. Through the encoding operation, an optimal intra-prediction mode for a PU of size 2N×2N can be derived. This optimal intra-prediction mode can be the intra-prediction mode that incurs the minimum rate-distortion cost when encoding a PU of size 2N×2N among the plurality of intra-prediction modes that can be used by encoding device 100.

[0462] Furthermore, the encoding device 100 can sequentially perform encoding operations on each PU obtained by performing N×N partitioning. Here, the encoding operation can be an operation of encoding the PU in each of a plurality of intra-prediction modes that can be used by the encoding device 100. Through the encoding operation, the optimal intra-prediction mode for a PU of size N×N can be derived. This optimal intra-prediction mode can be the intra-prediction mode that produces the minimum rate-distortion cost when encoding a PU of size N×N among the plurality of intra-prediction modes that can be used by the encoding device 100.

[0463] The encoding device 100 can determine which of the PUs, one of size 2N×2N and one of size N×N, will be encoded based on a comparison between the rate-distortion cost of a PU of size 2N×2N and the rate-distortion cost of a PU of size N×N.

[0464] A single CU can be partitioned into one or more PUs, and a PU can be partitioned into multiple PUs.

[0465] For example, when a single PU is partitioned into four PUs, the horizontal and vertical dimensions of each of the four PUs created through partitioning can be half the horizontal and vertical dimensions of the original PU. When a 32×32 PU is partitioned into four PUs, the size of each of the four partitioned PUs can be 16×16. When a single PU is partitioned into four PUs, the PU can be considered to have been partitioned in a quadtree structure.

[0466] For example, when a single PU is partitioned into two PUs, the horizontal or vertical dimension of each of the two resulting PUs can be half the horizontal or vertical dimension of the original PU. When a 32×32 PU is vertically partitioned into two PUs, the size of each of the two partitioned PUs can be 16×32. When a 32×32 PU is horizontally partitioned into two PUs, the size of each of the two partitioned PUs can be 32×16. When a single PU is partitioned into two PUs, the PU can be considered to have been partitioned in a binary tree structure.

[0467] Figure 5 This is a diagram showing the form of a transformation unit that can be included in an encoding unit.

[0468] A transform unit (TU) can be a basic unit in a CU used for processes such as transform, quantization, inverse transform, dequantization, entropy coding, and entropy decoding.

[0469] The TU can be square or rectangular. The shape of the TU can be determined based on the size and / or shape of the CU.

[0470] Within a CU partitioned from an LCU, CUs that are no longer designated as CUs can be partitioned into one or more TUs. Here, the partitioning structure of a TU can be a quadtree structure. For example, ... Figure 5 As shown, a single CU 510 can be partitioned once or more according to a quadtree structure. Through this partitioning, a single CU 510 can be composed of TUs of various sizes.

[0471] The CU can be considered to be recursively partitioned when a single CU is partitioned two or more times. Through partitioning, a single CU can be composed of transformation units (TUs) of various sizes.

[0472] Alternatively, a single CU can be divided into one or more TUs based on the number of vertical and / or horizontal lines dividing the CU.

[0473] The CU can be divided into symmetrical TUs or asymmetrical TUs. To divide into asymmetrical TUs, information about the size and / or shape of each TU can be transmitted from the encoding device 100 to the decoding device 200 via signals. Alternatively, the size and / or shape of each TU can be derived from the information about the size and / or shape of the CU.

[0474] A CU may not be classified as a TU. When a CU is not classified as a TU, the dimensions of the CU and the TU may be equal.

[0475] A single CU can be partitioned into one or more TUs, and a TU can be partitioned into multiple TUs.

[0476] For example, when a single TU is partitioned into four TUs, the horizontal and vertical dimensions of each of the four TUs generated by the partitioning can be half the horizontal and vertical dimensions of the original TU. When a TU of size 32×32 is partitioned into four TUs, the size of each of the four partitioned TUs can be 16×16. When a single TU is partitioned into four TUs, the TU can be considered to have been partitioned in a quadtree structure.

[0477] For example, when a single TU is partitioned into two TUs, the horizontal or vertical dimension of each of the two resulting TUs can be half the horizontal or vertical dimension of the original TU. When a TU of size 32×32 is vertically partitioned into two TUs, the size of each of the two partitioned TUs can be 16×32. When a TU of size 32×32 is horizontally partitioned into two TUs, the size of each of the two partitioned TUs can be 32×16. When a single TU is partitioned into two TUs, the TU can be considered to have been partitioned in a binary tree structure.

[0478] Can be with Figure 5 The different methods shown illustrate how CUs are divided.

[0479] For example, a single CU can be divided into three CUs. The horizontal or vertical dimensions of the three CUs generated by the division can be 1 / 4, 1 / 2, and 1 / 4 of the horizontal or vertical dimensions of the original CU before the division, respectively.

[0480] For example, when a 32×32 CU is vertically divided into three CUs, the resulting three CUs can have sizes of 8×32, 16×32, and 8×32, respectively. In this way, when a single CU is divided into three CUs, the CU can be considered to be divided in the form of a ternary tree.

[0481] One of the exemplary partitioning forms (i.e., quadtree partitioning, binary tree partitioning, and ternary tree partitioning) can be applied to the partitioning of the CU, and multiple partitioning schemes can be combined together for the partitioning of the CU. Here, the combination of multiple partitioning schemes is referred to as "composite tree partitioning".

[0482] Figure 6 This shows the block division based on the example.

[0483] During video encoding and / or decoding, such as Figure 6 As shown, the target block can be divided. For example, the target block can be a CU.

[0484] For the partitioning of the target block, an indicator indicating the partitioning information can be sent from the encoding device 100 to the decoding device 200 by a signal. The partitioning information can be information indicating how the target block is partitioned.

[0485] The partitioning information can be one or more of the following: a partitioning flag (hereinafter referred to as "split_flag"), a quad-binary flag (hereinafter referred to as "QB_flag"), a quadtree flag (hereinafter referred to as "quadtree_flag"), a binary tree flag (hereinafter referred to as "binarytree_flag"), and a binary type flag (hereinafter referred to as "Btype_flag").

[0486] The "split_flag" can be a flag indicating whether a block has been split. For example, a split_flag value of 1 indicates that the corresponding block has been split, while a split_flag value of 0 indicates that the corresponding block has not been split.

[0487] The "QB_flag" can be a flag indicating whether the block is partitioned into a quadtree or a binary tree. For example, a QB_flag value of 0 indicates that the block is partitioned into a quadtree, while a QB_flag value of 1 indicates that the block is partitioned into a binary tree. Alternatively, a QB_flag value of 0 indicates that the block is partitioned into a binary tree, while a QB_flag value of 1 indicates that the block is partitioned into a quadtree.

[0488] The "quadtree_flag" can be a flag indicating whether the block is partitioned as a quadtree. For example, a quadtree_flag value of 1 indicates that the block is partitioned as a quadtree, while a quadtree_flag value of 0 indicates that the block is not partitioned as a quadtree.

[0489] The "binarytree_flag" can be a flag indicating whether the block is partitioned as a binary tree. For example, a binarytree_flag value of 1 indicates that the block is partitioned as a binary tree, while a binarytree_flag value of 0 indicates that the block is not partitioned as a binary tree.

[0490] The `Btype_flag` can be a flag indicating which of the vertical or horizontal partitions corresponds to the partitioning direction when a block is divided in a binary tree format. For example, a `Btype_flag` value of 0 indicates that the block is partitioned horizontally, and a `Btype_flag` value of 1 indicates that the block is partitioned vertically. Alternatively, a `Btype_flag` value of 0 indicates that the block is partitioned vertically, and a `Btype_flag` value of 1 indicates that the block is partitioned horizontally.

[0491] For example, it can be derived by sending at least one of quadtree_flag, binarytree_flag, and Btype_flag using signals. Figure 6 The block division information is shown in Table 1 below.

[0492] Table 1

[0493] For example, it can be derived by sending at least one of split_flag, QB_flag, and Btype_flag using signals. Figure 6 The block division information is shown in Table 2 below.

[0494] Table 2

[0495] The partitioning method may be limited to quadtrees or binary trees depending on the size and / or shape of the blocks. When this restriction is applied, the `split_flag` may be a flag indicating whether the blocks are partitioned in a quadtree or a binary tree format. The size and shape of the blocks can be derived from the block depth information, and the depth information can be transmitted from the encoding device 100 to the decoding device 200 by a signal.

[0496] When the block size falls within a certain range, it is possible to partition in quadtree form only. For example, the certain range can be defined by at least one of the maximum block size and the minimum block size that can be partitioned in quadtree form only.

[0497] Information indicating the maximum and minimum block sizes that can be partitioned only in quadtree form can be transmitted from the encoding device 100 to the decoding device 200 via a bitstream. Furthermore, this information can be transmitted via a signal for at least one of the units such as video, sequence, picture, parameter, parallel block group, and strip (or segment).

[0498] Alternatively, the maximum block size and / or minimum block size can be fixed sizes predefined by the encoding device 100 and the decoding device 200. For example, when the block size is greater than 64×64 and less than 256×256, it is possible to partition the block only in the form of a quadtree. In this case, the split_flag can be a flag indicating whether to perform partitioning in the form of a quadtree.

[0499] When the size of a block is larger than the maximum size of a transform block, it is possible to partition it only in the form of a quadtree. Here, the sub-blocks generated by partitioning can be at least one of CU and TU.

[0500] In this case, split_flag can be a flag indicating whether the CU is partitioned in the form of a quadtree.

[0501] When the size of a block falls within a certain range, it is possible to partition it using only a binary tree or a ternary tree. For example, the specific range can be defined by at least one of the maximum block size and the minimum block size that allows partitioning using only a binary tree or a ternary tree.

[0502] Information indicating the maximum and / or minimum block size, which can be partitioned in a binary tree or a ternary tree manner, can be transmitted from the encoding device 100 to the decoding device 200 via a bitstream. Furthermore, this information can be transmitted via a signal for at least one of the units such as sequences, frames, and stripes (or segments).

[0503] Alternatively, the maximum block size and / or minimum block size can be fixed sizes predefined by the encoding device 100 and the decoding device 200. For example, when the block size is greater than 8×8 and less than 16×16, it is possible to partition in binary tree form only. In this case, split_flag can be a flag indicating whether to perform partitioning in binary tree or ternary tree form.

[0504] The above description of partitioning in the form of a quadtree can be applied equally to binary and / or ternary tree forms.

[0505] The partitioning of a block may be limited by previous partitioning. For example, when a block is partitioned in a specific binary tree form and multiple sub-blocks are generated from said partition, each sub-block may be further partitioned only in that specific tree form. Here, the specific tree form can be at least one of a binary tree form, a ternary tree form, and a quadtree form.

[0506] When the horizontal or vertical dimensions of a partition block are dimensions that cannot be further subdivided, the aforementioned indicator does not need to be sent.

[0507] Figure 7 This is a diagram illustrating an embodiment used to explain the intra-frame prediction process.

[0508] from Figure 7 The radially extending arrow from the center of the diagram indicates the prediction direction of the intra-prediction mode. Furthermore, the numbers appearing near the arrows indicate examples of mode values ​​assigned to the intra-prediction mode or its prediction direction.

[0509] exist Figure 7 In the diagram, number 0 can represent the planar mode as a non-directional intra-prediction mode. Number 1 can represent the DC mode as a non-directional intra-prediction mode.

[0510] Intra-frame coding and / or decoding can be performed using reference samples from neighboring blocks of the target block. A neighboring block can be a reconstructed neighboring block. Reference samples can represent neighboring samples.

[0511] For example, intra-frame encoding and / or decoding can be performed using the values ​​of reference samples included in the reconstructed neighboring blocks or the encoding parameters of the reconstructed neighboring blocks.

[0512] Encoding device 100 and / or decoding device 200 can generate a prediction block by performing intra-frame prediction on the target block based on information about samples in the target image. When intra-frame prediction is performed, encoding device 100 and / or decoding device 200 can generate a prediction block for the target block by performing intra-frame prediction based on information about samples in the target image. When intra-frame prediction is performed, encoding device 100 and / or decoding device 200 can perform directional prediction and / or non-directional prediction based on at least one reconstructed reference sample.

[0513] A prediction block can be a block generated as a result of performing intra-frame prediction. A prediction block can correspond to at least one of CU, PU, ​​and TU.

[0514] The cells of the prediction block may have a size corresponding to at least one of CU, PU, ​​and TU. The prediction block may have a square shape with a size of 2N×2N or N×N. The size N×N may include sizes such as 4×4, 8×8, 16×16, 32×32, 64×64, etc.

[0515] Alternatively, the prediction block can be a square block with a size of 2×2, 4×4, 8×8, 16×16, 32×32, 64×64, etc., or a rectangular block with a size of 2×8, 4×8, 2×16, 4×16, 8×16, etc.

[0516] Intra-prediction can be performed using intra-prediction modes for the target block. The number of intra-prediction modes that a target block can have can be a predefined fixed value, or it can be a value determined differently based on the attributes of the prediction block. For example, the attributes of the prediction block can include the size of the prediction block, the type of the prediction block, etc. In addition, the attributes of the prediction block can indicate the coding parameters used for the prediction block.

[0517] For example, the number of intra-prediction modes can be fixed at N, regardless of the size of the prediction block. Alternatively, the number of intra-prediction modes can be, for example, 3, 5, 9, 17, 34, 35, 36, 65, 67, or 95.

[0518] Intra-frame prediction mode can be either non-directional or directional.

[0519] For example, intra-frame prediction modes may include... Figure 7 The numbers 0 to 66 shown correspond to two non-directional modes and 65 directional modes.

[0520] For example, when using a specific intra-prediction method, the intra-prediction mode may include... Figure 7 The numbers -14 to 80 shown correspond to the two non-directional patterns and 93 directional patterns.

[0521] The two non-directional modes may include DC mode and planar mode.

[0522] Directional patterns can be prediction patterns with a specific direction or angle. Directional patterns can also be called "angle patterns".

[0523] An intra-prediction mode can be represented by at least one of a mode number, a mode value, a mode angle, and a mode direction. In other words, the terms "(mode) number of intra-prediction mode", "(mode) value of intra-prediction mode", "(mode) angle of intra-prediction mode", and "(mode) direction of intra-prediction mode" can be used to have the same meaning and can be used interchangeably with each other.

[0524] The number of intra-prediction modes can be M. The value of M can be 1 or greater. In other words, the number of intra-prediction modes can be M, where M includes the number of non-directional modes and the number of directional modes.

[0525] The number of intra-prediction modes can be fixed at M, regardless of the block size and / or color components. For example, the number of intra-prediction modes can be fixed at either 35 or 67, regardless of the block size.

[0526] Alternatively, the number of intra-frame prediction modes may vary depending on the shape, size, and / or type of color components of the block.

[0527] For example, in Figure 7 In the diagram, the direction prediction pattern shown by the dashed line can only be applied to the prediction of non-square blocks.

[0528] For example, the larger the block size, the more intra-prediction modes are available. Alternatively, the larger the block size, the fewer intra-prediction modes are available. When the block size is 4×4 or 8×8, the number of intra-prediction modes can be 67. When the block size is 16×16, the number of intra-prediction modes can be 35. When the block size is 32×32, the number of intra-prediction modes can be 19. When the block size is 64×64, the number of intra-prediction modes can be 7.

[0529] For example, the number of intra-prediction modes can vary depending on whether the color component is a luma signal or a chrominance signal. Alternatively, the number of intra-prediction modes corresponding to the luma component block can be greater than the number of intra-prediction modes corresponding to the chrominance component block.

[0530] For example, in the vertical mode with a mode value of 50, prediction can be performed along the vertical direction based on the pixel values ​​of the reference sample. Similarly, in the horizontal mode with a mode value of 18, prediction can be performed along the horizontal direction based on the pixel values ​​of the reference sample.

[0531] Even in directional modes other than those described above, the encoding device 100 and the decoding device 200 can still perform intra-frame prediction on the target unit using reference samples based on the angle corresponding to the directional mode.

[0532] Intra-prediction modes located to the right of the vertical mode can be called "vertical-right mode". Intra-prediction modes located below the horizontal mode can be called "horizontal-bottom mode". For example, in Figure 7 In the frame prediction mode, the mode value being one of 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, and 66 can be a vertical-right mode. The mode value being one of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, and 17 can be a horizontal-downward mode.

[0533] Non-directional modes can include DC mode and planar mode. For example, the value for DC mode can be 1, and the value for planar mode can be 0.

[0534] Orientation modes can include angle modes. Among the various intra-frame prediction modes, all modes except DC mode and planar mode can be orientation modes.

[0535] When the intra-frame prediction mode is DC mode, a prediction block can be generated based on the average pixel values ​​of multiple reference pixels. For example, the pixel values ​​of the prediction block can be determined based on the average pixel values ​​of multiple reference pixels.

[0536] The number of intra-prediction modes and the mode values ​​of each intra-prediction mode described above are merely exemplary. The number of intra-prediction modes and the mode values ​​of each intra-prediction mode described above may be defined differently depending on the embodiment, implementation, and / or requirements.

[0537] To perform intra-frame prediction on a target block, a step can be performed to check whether samples included in the reconstructed neighboring blocks can be used as reference samples for the target block. When there are samples among the samples in the neighboring blocks that cannot be used as reference samples for the target block, a value generated by interpolation and / or duplication using at least one sample value from the samples included in the reconstructed neighboring blocks can replace the sample value of the sample that cannot be used as a reference sample. When a value generated by duplication and / or interpolation replaces the sample value of an existing sample, that sample can be used as a reference sample for the target block.

[0538] When using intra-frame prediction, filters can be applied to at least one of the reference samples and the prediction samples based on the size of the target block and at least one of the intra-frame prediction modes.

[0539] The type of filter to be applied to at least one of the reference sample and the prediction sample can vary based on at least one of the intra-prediction mode of the target block, the size of the target block, and the shape of the target block. The filter type can be classified based on one or more of the filter tap length, the value of the filter coefficients, and the filter strength. The filter tap length can represent the number of filter taps. Furthermore, the number of filter taps can represent the length of the filter.

[0540] When the intra-frame prediction mode is planar mode, the sample value of the predicted target block can be generated by weighting the top reference sample, left reference sample, upper right reference sample, and lower left reference sample of the target block according to the position of the predicted target sample in the prediction block.

[0541] When the intra-frame prediction mode is DC mode, the average of reference samples above and to the left of the target block can be used when generating the prediction block for the target block. Furthermore, filtering using the values ​​of the reference samples can be performed on specific rows or columns within the target block. The specific row can be one or more upper rows adjacent to the reference samples. The specific column can be one or more left columns adjacent to the reference samples.

[0542] When the intra-frame prediction mode is directional mode, the top reference sample, left reference sample, top right reference sample, and / or bottom left reference sample of the target block can be used to generate the prediction block.

[0543] To generate the above predicted samples, real-number-based interpolation can be performed.

[0544] The intra-prediction mode of the target block can be predicted from the intra-prediction modes of neighboring blocks adjacent to the target block, and the information used for prediction can be entropy encoded / entropy decoded.

[0545] For example, when the intra-prediction modes of the target block and neighboring blocks are the same, a predefined flag can be used to signal that the intra-prediction modes of the target block and neighboring blocks are the same.

[0546] For example, an indicator can be sent to indicate 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.

[0547] When the intra-prediction modes of the target block and neighboring blocks are different from each other, entropy coding and / or entropy decoding can be used to encode and / or decode information about the intra-prediction mode of the target block.

[0548] Figure 8 This is a diagram showing the reference samples used in the intra-frame prediction process.

[0549] The reconstruction reference points used for intra-frame prediction of the target block may include the lower left reference point, the left reference point, the upper left corner reference point, the upper reference point, and the upper right reference point.

[0550] For example, a left reference sample may represent a reconstructed reference pixel adjacent to the left side of the target block. A top reference sample may represent a reconstructed reference pixel adjacent to the top of the target block. A top-left reference sample may represent a reconstructed reference pixel located at the top-left corner of the target block. A bottom-left reference sample may represent a reference sample located below the left-side sample line, which is the same line as the left-side sample line formed by the left reference samples. A top-right reference sample may represent a reference sample located to the right of the upper sample line, which is the same line as the upper sample line formed by the upper reference samples.

[0551] When the size of the target block is N×N, the number of the lower left reference point, the left reference point, the upper reference point, and the upper right reference point can all be N.

[0552] A prediction block can be generated by performing intra-frame prediction on the target block. The process of generating a prediction block may include determining the values ​​of the pixels in the prediction block. The target block and the prediction block can have the same size.

[0553] The reference sample used for intra-prediction of the target block can be changed depending on the intra-prediction mode of the target block. The direction of the intra-prediction mode can represent the dependency between the reference sample and the pixels of the prediction block. For example, the value of a specified reference sample can be used as the value of one or more specified pixels in the prediction block. In this case, the specified reference sample and the one or more specified pixels in the prediction block can be samples and pixels located on a straight line along the direction of the intra-prediction mode. In other words, the value of the specified reference sample can be copied as the value of a pixel located in the opposite direction to the direction of the intra-prediction mode. Alternatively, the value of a pixel in the prediction block can be the value of a reference sample located in the direction of the intra-prediction mode relative to the pixel's position.

[0554] In the example, when the intra-prediction mode of the target block is vertical, the upper reference sample can be used for intra-prediction. When the intra-prediction mode is vertical, the value of a pixel in the prediction block can be the value of a reference sample vertically above that pixel. Therefore, the upper reference sample adjacent to the top of the target block can be used for intra-prediction. Furthermore, the value of a pixel in a row of the prediction block can be the same as the value of a pixel at the upper reference sample.

[0555] In the example, when the intra-prediction mode of the target block is horizontal, the left reference sample can be used for intra-prediction. When the intra-prediction mode is horizontal, the value of a pixel in the prediction block can be the value of a reference sample horizontally to the left of that pixel. Therefore, the left reference sample adjacent to the left side of the target block can be used for intra-prediction. Furthermore, the value of a pixel in a column of the prediction block can be the same as the value of the pixel in the left reference sample.

[0556] In the example, when the mode value of the intra-prediction mode for the current block is 34, at least some of the left reference samples, the top-left reference sample, and the top reference sample can be used for intra-prediction. When the mode value of the intra-prediction mode is 34, the value of a pixel in the prediction block can be the value of a reference sample located diagonally at the top-left corner of that pixel.

[0557] Furthermore, in the case of an intra-prediction mode with mode values ​​ranging from 52 to 66, at least a portion of the upper right reference samples can be used for intra-prediction.

[0558] Furthermore, in the case of intra-prediction modes with mode values ​​ranging from 2 to 17, at least a portion of the lower left reference samples can be used for intra-prediction.

[0559] Furthermore, in the case of intra-prediction modes with mode values ​​ranging from 19 to 49, the upper left reference sample can be used for intra-prediction.

[0560] The number of reference samples used to determine the pixel value of a pixel in the prediction block can be 1, 2, or more.

[0561] As described above, the pixel value of a pixel in a prediction block can be determined based on the pixel's position and the position of a reference sample indicated by the direction of the intra-prediction mode. When both the pixel's position and the position of the reference sample indicated by the direction of the intra-prediction mode are integer positions, the value of a reference sample indicated by the integer position can be used to determine the pixel value of the pixel in the prediction block.

[0562] When the pixel position and the position of the reference sample indicated by the direction of the intra-prediction mode are not integer positions, an interpolated reference sample can be generated based on the two reference samples closest to that reference sample position. The value of the interpolated reference sample can be used to determine the pixel value of the pixel in the prediction block. In other words, when the pixel position in the prediction block and the position of the reference sample indicated by the direction of the intra-prediction mode indicate the position between two reference samples, an interpolation based on the values ​​of those two samples can be generated.

[0563] The predicted block generated by prediction may differ from the original target block. In other words, there may be prediction errors, which are the differences between the target block and the predicted block, and there may also be prediction errors between pixels in the target block and pixels in the predicted block.

[0564] In the following text, the terms “difference,” “error,” and “residual” are used to have the same meaning and are interchangeable.

[0565] For example, in the case of intra-frame prediction, the greater the distance between the pixels of the predicted block and the reference sample, the greater the potential prediction error. This prediction error can lead to discontinuities between the generated predicted block and its neighboring blocks.

[0566] To reduce prediction error, filtering operations can be used for prediction blocks. These filtering operations can be configured to adaptively apply filters to regions within the prediction block that are considered to have large prediction errors. For example, regions considered to have large prediction errors could be the boundaries of the prediction block. Furthermore, the regions within the prediction block considered to have large prediction errors can vary depending on the intra-prediction mode, and the characteristics of the filters can also vary depending on the intra-prediction mode.

[0567] likeFigure 8 As shown, for intra-frame prediction of the target block, at least one of reference lines 0 to 3 can be used.

[0568] exist Figure 8 Each reference line in the code can indicate a reference point line that includes one or more reference points. When the reference line number is smaller, it can indicate a reference point line that is closer to the target block.

[0569] Samples in fragments A and F can be obtained by padding instead of from reconstructed neighboring blocks, wherein the padding uses the samples from fragments B and E that are closest to the target block.

[0570] An index information indicating the reference sample lines to be used for intra-frame prediction of the target block can be transmitted using a signal. The index information can indicate which of a plurality of reference sample lines will be used for intra-frame prediction of the target block. For example, the index information can have a value corresponding to any one of 0 to 3.

[0571] When the upper boundary of the target block is the boundary of the CTU, only reference sample line 0 can be available. Therefore, in this case, index information does not need to be sent. When additional reference sample lines besides reference sample line 0 are used, filtering of the prediction block, which will be described later, is not required.

[0572] In the case of intra-frame prediction between colors, the predicted block of the target block of the second color component can be generated based on the corresponding reconstructed block of the first color component.

[0573] For example, the first color component can be the luminance component, and the second color component can be the chromaticity component.

[0574] To perform inter-color intra-frame prediction, the parameters of a linear model between the first and second color components can be derived based on a template.

[0575] The template may include a reference point above the target block (upper reference point) and / or a reference point to the left of the target block (left reference point), and may include the upper reference point and / or left reference point of the reconstructed block of the first color component corresponding to the reference point.

[0576] For example, the following values ​​can be used to derive the parameters of a linear model: 1) the value of the sample point of the first color component with the maximum value among the samples in the template, 2) the value of the sample point of the second color component corresponding to the sample point of the first color component, 3) the value of the sample point of the first color component with the minimum value among the samples in the template, and 4) the value of the sample point of the second color component corresponding to the sample point of the first color component.

[0577] When exporting the parameters of a linear model, the predicted block of the target block can be generated by applying the corresponding reconstructed block to the linear model.

[0578] Depending on the image format, subsampling can be performed on samples adjacent to the reconstructed block of the first color component and on the corresponding reconstructed block of the first color component. For example, when one sample of the second color component corresponds to four samples of the first color component, a corresponding sample can be calculated by subsampling the four samples of the first color component. When subsampling is performed, the parameters of the linear model and inter-color intra-frame prediction can be performed based on the subsampled corresponding sample.

[0579] In intra-frame prediction mode, information about whether to perform inter-color intra-frame prediction and / or the range of templates can be sent by signaling.

[0580] The target block can be divided into two or four sub-blocks in the horizontal and / or vertical directions.

[0581] Sub-blocks generated by partitioning can be reconstructed sequentially. That is, when intra-prediction is performed on each sub-block, a sub-prediction block for that sub-block can be generated. Furthermore, when inverse quantization and / or inverse transform is performed on each sub-block, a sub-residual block for the corresponding sub-block can be generated. Reconstructed sub-blocks can be generated by adding the sub-prediction blocks to the sub-residual blocks. The reconstructed sub-blocks can be used as reference samples for intra-prediction of sub-blocks with the next higher priority.

[0582] A sub-block can be a block containing a specific number (e.g., 16) or more samples. For example, when the target block is an 8×4 block or a 4×8 block, the target block can be divided into two sub-blocks. Furthermore, when the target block is a 4×4 block, it cannot be divided into sub-blocks. When the target block has another size, it can be divided into four sub-blocks.

[0583] Signals can be used to send information about whether to perform intra-frame prediction based on these sub-blocks and / or about the partitioning direction (horizontal or vertical).

[0584] This sub-block-based intra-prediction can be restricted so that it is performed only when reference sample line 0 is used. When performing sub-block-based intra-prediction, filtering of the prediction block, which will be described below, may not be performed.

[0585] The final prediction block can be generated by filtering the prediction block generated via intra-frame prediction.

[0586] Filtering can be performed by applying specific weights to the target sample, left reference sample, top reference sample, and / or top-left reference sample, which are the targets to be filtered.

[0587] The weights and / or reference samples (e.g., the range of reference samples, the location of reference samples, etc.) used for filtering can be determined based on at least one of the block size, intra-frame prediction mode, and the location of the target filter sample in the prediction block.

[0588] For example, filtering can be performed only in specific intra-frame prediction modes (e.g., DC mode, planar mode, vertical mode, horizontal mode, diagonal mode, and / or adjacent diagonal mode).

[0589] Adjacent diagonal patterns can be patterns with numbers obtained by adding k to the diagonal pattern's number, or patterns with numbers obtained by subtracting k from the diagonal pattern's number. In other words, the number of an adjacent diagonal pattern can be the sum of the diagonal pattern's number and k, or the difference between the diagonal pattern's number and k. For example, k can be a positive integer of 8 or less.

[0590] The intra prediction mode of the target block can be derived using the intra prediction modes of neighboring blocks that appear around the target block, and this derived intra prediction mode can be entropy encoded and / or entropy decoded.

[0591] For example, when the intra prediction mode of the target block is the same as that of the neighboring blocks, specific flag information can be used to signal information indicating that the intra prediction mode of the target block is the same as that of the neighboring blocks.

[0592] Furthermore, for example, an indicator information of neighboring blocks whose intra-prediction modes are the same as those of the target block can be transmitted using signals.

[0593] For example, when the intra prediction mode of the target block is different from that of the neighboring blocks, entropy coding and / or entropy decoding can be performed on the information about the intra prediction mode of the target block by performing entropy coding and / or entropy decoding based on the intra prediction modes of the neighboring blocks.

[0594] Figure 9 This is a diagram illustrating an embodiment used to explain the inter-frame prediction process.

[0595] Figure 9 The rectangles shown can represent images (or screens). Furthermore, in... Figure 9 In the image, arrows indicate the prediction direction. An arrow pointing from the first frame to the second frame indicates that the second frame references the first frame. In other words, each image can be encoded and / or decoded based on the prediction direction.

[0596] Images can be classified into intra-frame frames (I-frames), single-predictive or predictive-coded frames (P-frames), and double-predictive or double-predictive-coded frames (B-frames) based on their encoding type. Each frame can be encoded and / or decoded according to its encoding type.

[0597] When the target image to be encoded is an I-frame, the target image can be encoded using the data contained in the image itself without inter-frame prediction referencing other images. For example, an I-frame can be encoded solely via intra-frame prediction.

[0598] When the target image is a P-frame, it can be encoded using inter-frame prediction with reference frames existing in one direction. Here, the one direction can be a forward direction or a backward direction.

[0599] When the target image is a B-frame, the image can be encoded via inter-frame prediction using reference frames present in both directions, or via inter-frame prediction using reference frames present in one of the forward and backward directions. Here, the two directions can be the forward and backward directions.

[0600] P-frames and B-frames that are encoded and / or decoded using reference frames can be considered as images using inter-frame prediction.

[0601] The following will describe in detail the inter-frame prediction in inter-frame mode according to the embodiments.

[0602] Reference images and motion information can be used to perform inter-frame prediction or motion compensation.

[0603] In inter-frame mode, encoding device 100 may perform inter-frame prediction and / or motion compensation on the target block. Decoding device 200 may perform inter-frame prediction and / or motion compensation on the target block corresponding to the inter-frame prediction and / or motion compensation performed by encoding device 100.

[0604] Motion information of the target block can be derived separately by the encoding device 100 and the decoding device 200 during inter-frame prediction. Motion information can be derived using the motion information of reconstructed neighboring blocks, the motion information of the col block, and / or the motion information of blocks adjacent to the col block.

[0605] For example, encoding device 100 or decoding device 200 can perform prediction and / or motion compensation by using motion information of spatial candidates and / or temporal candidates as motion information for a target block. The target block may represent a PU and / or a PU partition.

[0606] Spatial candidates can be reconstructed blocks that are spatially adjacent to the target block.

[0607] The time candidate can be a reconstructed block that corresponds to the target block in a previously reconstructed co-location frame (col frame).

[0608] In inter-frame prediction, the encoding device 100 and the decoding device 200 can improve encoding efficiency and decoding efficiency by utilizing motion information from spatial candidates and / or temporal candidates. The motion information from spatial candidates can be referred to as "spatial motion information." The motion information from temporal candidates can be referred to as "temporal motion information."

[0609] Below, the motion information of spatial candidates can be the motion information of PUs including spatial candidates. The motion information of temporal candidates can be the motion information of PUs including temporal candidates. The motion information of candidate blocks can be the motion information of PUs including candidate blocks.

[0610] Inter-frame prediction can be performed using a reference frame.

[0611] The reference image can be at least one of an image preceding or following the target image. The reference image can be an image used for prediction of the target block.

[0612] In inter-frame prediction, regions within a reference frame can be specified using a reference frame index (or refIdx) used to indicate the reference frame, motion vectors that will be described subsequently, and so on. Here, the region specified in the reference frame can indicate a reference block.

[0613] Inter-frame prediction can select a reference frame, and can also select a reference block corresponding to the target block from the reference frame. In addition, inter-frame prediction can use the selected reference block to generate a prediction block for the target block.

[0614] Motion information can be derived by each of the encoding device 100 and the decoding device 200 during inter-frame prediction.

[0615] Spatial candidates can be 1) blocks that exist in the target frame, 2) blocks that have been previously reconstructed via encoding and / or decoding, and 3) blocks that are adjacent to or located at the corner of the target block. Here, a "block located at the corner of the target block" can be a block that is vertically adjacent to a horizontally adjacent neighboring block, or a block that is horizontally adjacent to a vertically adjacent neighboring block. Furthermore, "block located at the corner of the target block" can have the same meaning as "block adjacent to the corner of the target block." The meaning of "block located at the corner of the target block" can be included within the meaning of "block adjacent to the target block."

[0616] For example, a spatial candidate can be a reconstruction block located to the left of the target block, a reconstruction block located above the target block, a reconstruction block located at the lower left corner of the target block, a reconstruction block located at the upper right corner of the target block, or a reconstruction block located at the upper left corner of the target block.

[0617] Each of the encoding device 100 and the decoding device 200 can identify a block existing in the col frame at a spatial position corresponding to the target block. The position of the target block in the target frame and the position of the identified block in the col frame can correspond to each other.

[0618] Each of the encoding device 100 and the decoding device 200 can identify a col block existing at a predefined relevant location for the identified block as a time candidate. The predefined relevant location can be a location existing inside and / or outside the identified block.

[0619] For example, a col block can include a first col block and a second col block. When the coordinates of the identified block are (xP, yP) and the size of the identified block is represented by (nPSW, nPSH), the first col block can be the block located at coordinates (xP+nPSW, yP+nPSH). The second col block can be the block located at coordinates (xP+(nPSW>>1), yP+(nPSH>>1)). When the first col block is unavailable, the second col block can be used selectively.

[0620] The motion vector of the target block can be determined based on the motion vector of the col block. Each of the encoding device 100 and the decoding device 200 can scale the motion vector of the col block. The scaled motion vector of the col block can be used as the motion vector of the target block. Furthermore, the motion vectors of motion information for time candidates stored in a list can also be scaled motion vectors.

[0621] The ratio of the motion vector of the target block to the motion vector of the col block can be the same as the ratio of the first time distance to the second time distance. The first time distance can be the distance between the reference frame and the target frame of the target block. The second time distance can be the distance between the reference frame and the col frame of the col block.

[0622] The scheme used to derive motion information can be varied depending on the inter-frame prediction mode of the target block. For example, inter-frame prediction modes applied to inter-frame prediction may include Advanced Motion Vector Prediction Factor (AMVP) mode, merge mode, skip mode, merge mode with motion vector difference, sub-block merge mode, triangular partitioning mode, inter-frame / intra-frame combined prediction mode, affine inter-frame mode, and current frame reference mode. The merge mode can also be called "motion merge mode." Each mode will be described in detail below.

[0623] 1) AMVP mode When using AMVP mode, the encoding device 100 can search for similar blocks in the neighborhood of the target block. The encoding device 100 can obtain a predicted block by performing a prediction on the target block using the motion information of the found similar blocks. The encoding device 100 can encode the residual block, which is the difference between the target block and the predicted block.

[0624] 1-1) Creation of a list of prediction motion vector candidates When the AMVP mode is used as the prediction mode, each of the encoding device 100 and the decoding device 200 can use spatial candidate motion vectors, temporal candidate motion vectors, and zero vectors to create a list of prediction motion vector candidates. The prediction motion vector candidate list may include one or more prediction motion vector candidates. At least one of the spatial candidate motion vectors, temporal candidate motion vectors, and zero vectors can be determined and used as a prediction motion vector candidate.

[0625] In the following text, the terms “predicted motion vector (candidate)” and “motion vector (candidate)” can be used to have the same meaning and can be used interchangeably.

[0626] In the following text, the terms “predicted motion vector candidate” and “AMVP candidate” can be used to have the same meaning and can be used interchangeably.

[0627] In the following text, the terms “predicted motion vector candidate list” and “AMVP candidate list” can be used to have the same meaning and can be used interchangeably.

[0628] Spatial candidates can include reconstructed spatial neighbor blocks. In other words, the motion vectors of the reconstructed neighbor blocks can be referred to as "spatial prediction motion vector candidates".

[0629] A temporal candidate can include the col block and the blocks adjacent to the col block. In other words, the motion vector of the col block or the motion vector of the blocks adjacent to the col block can be called a "temporal prediction motion vector candidate".

[0630] The zero vector can be a (0,0) motion vector.

[0631] The predicted motion vector candidate can be a motion vector predictor used to predict the motion vector. Furthermore, in the encoding device 100, each predicted motion vector candidate can be an initial search position for the motion vector.

[0632] 1-2) Search for a motion vector using the list of prediction motion vector candidates Encoding device 100 can use a list of predicted motion vector candidates to determine, within a search range, the motion vectors that will be used to encode the target block. Furthermore, encoding device 100 can determine, from among the predicted motion vector candidates present in the list of predicted motion vector candidates, the predicted motion vector candidates that will be used as the target block.

[0633] The motion vector used to encode the target block can be a motion vector that can be encoded at the minimum cost.

[0634] In addition, the encoding device 100 can determine whether to use the AMVP mode to encode the target block.

[0635] 1-3) Transmission of inter prediction information Encoding device 100 can generate a bitstream that includes inter-frame prediction information required for inter-frame prediction. Decoding device 200 can use the inter-frame prediction information of the bitstream to perform inter-frame prediction on the target block.

[0636] Inter-frame prediction information may include 1) mode information indicating whether the AMVP mode is used, 2) predicted motion vector index, 3) motion vector difference (MVD), 4) reference direction and 5) reference frame index.

[0637] In the following text, the terms “predicted motion vector index” and “AMVP index” can be used to have the same meaning and can be used interchangeably.

[0638] In addition, inter-frame prediction information may include residual signals.

[0639] When the mode information indicates that the AMVP mode is used, the decoding device 200 can obtain the predicted motion vector index, MVD, reference direction, and reference frame index from the bitstream through entropy decoding.

[0640] The predicted motion vector index indicates which of the predicted motion vector candidates included in the predicted motion vector candidate list will be used to predict the target block.

[0641] 1-4) Inter prediction in AMVP mode using inter prediction information The decoding device 200 can use the list of predicted motion vector candidates to derive predicted motion vector candidates, and can determine the motion information of the target block based on the derived predicted motion vector candidates.

[0642] The decoding device 200 can use the predicted motion vector index to determine a motion vector candidate for the target block from among the predicted motion vector candidates included in the predicted motion vector candidate list. The decoding device 200 can select the predicted motion vector candidate indicated by the predicted motion vector index from among the predicted motion vector candidates included in the predicted motion vector candidate list as the predicted motion vector for the target block.

[0643] Encoding device 100 can generate an entropy-coded predicted motion vector index by applying entropy coding to the predicted motion vector index, and can generate a bitstream including the entropy-coded predicted motion vector index. The entropy-coded predicted motion vector index can be transmitted from encoding device 100 to decoding device 200 via the bitstream. Decoding device 200 can extract the entropy-coded predicted motion vector index from the bitstream, and can obtain the predicted motion vector index by applying entropy decoding to the entropy-coded predicted motion vector index.

[0644] The motion vector actually used for inter-frame prediction of the target block may not match the predicted motion vector. MVD (Motion Vector Difference) can be used to indicate the difference between the actual motion vector used for inter-frame prediction of the target block and the predicted motion vector. The encoding device 100 can derive a predicted motion vector similar to the actual motion vector used for inter-frame prediction of the target block in order to use the smallest possible MVD.

[0645] Motion Vector Difference (MVD) can be the difference between the motion vector of the target block and the predicted motion vector. Encoding device 100 can compute the MVD and generate an entropy-coded MVD by applying entropy coding to the MVD. Encoding device 100 can generate a bitstream including the entropy-coded MVD.

[0646] The MVD can be sent from the encoding device 100 to the decoding device 200 via a bitstream. The decoding device 200 can extract the entropy-encoded MVD from the bitstream and obtain the MVD by applying entropy decoding to the entropy-encoded MVD.

[0647] The decoding device 200 can derive the motion vector of the target block by summing the MVD and the predicted motion vector. In other words, the motion vector of the target block derived by the decoding device 200 can be the sum of the MVD and the motion vector candidate.

[0648] Furthermore, the encoding device 100 can generate entropy-coded MVD resolution information by applying entropy coding to the calculated MVD resolution information, and can generate a bitstream including the entropy-coded MVD resolution information. The decoding device 200 can extract the entropy-coded MVD resolution information from the bitstream, and can obtain the MVD resolution information by applying entropy decoding to the entropy-coded MVD resolution information. The decoding device 200 can use the MVD resolution information to adjust the MVD resolution.

[0649] Additionally, the encoding device 100 can calculate the MVD based on an affine model. The decoding device 200 can derive the affine control motion vector of the target block from the sum of the MVD and the affine control motion vector candidates, and can use the affine control motion vector to derive the motion vector of the sub-blocks.

[0650] The reference direction can indicate a list of reference frames that will be used to predict the target block. For example, the reference direction can indicate one of reference frame list L0 and reference frame list L1.

[0651] The reference direction only indicates the list of reference frames that will be used to predict the target block, and does not necessarily mean that the direction of the reference frames is limited to the forward or backward direction. In other words, each of the reference frame lists L0 and L1 can include frames in the forward and / or backward directions.

[0652] A unidirectional reference direction can mean using a single reference screen list. A bidirectional reference direction can mean using two reference screen lists. In other words, the reference direction can indicate one of the following: using only reference screen list L0, using only reference screen list L1, or using both reference screen lists.

[0653] A reference frame index indicates a reference frame in a list of reference frames used to predict a target block. Encoding device 100 can generate an entropy-coded reference frame index by applying entropy coding to the reference frame index, and can generate a bitstream including the entropy-coded reference frame index. The entropy-coded reference frame index can be signaled from encoding device 100 to decoding device 200 via the bitstream. Decoding device 200 can extract the entropy-coded reference frame index from the bitstream, and can obtain the reference frame index by applying entropy decoding to the entropy-coded reference frame index.

[0654] When two reference frame lists are used to predict a target block, a single reference frame index and a single motion vector can be used for each of the reference frame lists. Furthermore, when two reference frame lists are used to predict a target block, two prediction blocks can be specified for the target block. For example, the (final) prediction block for the target block can be generated using the average or weighted sum of the two prediction blocks for the target block.

[0655] The motion vector of a target block can be derived by predicting the motion vector index, MVD, reference direction, and reference screen index.

[0656] The decoding device 200 can generate a predicted block for a target block based on the exported motion vectors and a reference frame index. For example, the predicted block can be a reference block indicated by the exported motion vectors in a reference frame indicated by the reference frame index.

[0657] Since the predicted motion vector index and MVD are encoded, while the motion vector of the target block itself is not encoded, the number of bits sent from the encoding device 100 to the decoding device 200 can be reduced, and the encoding efficiency can be improved.

[0658] For the target block, motion information from reconstructed neighboring blocks can be used. In certain inter-frame prediction modes, the encoding device 100 may not encode the actual motion information of the target block separately. Instead of encoding the motion information of the target block, additional information can be encoded, which enables the derivation of the target block's motion information using the motion information from reconstructed neighboring blocks. Because this additional information is encoded, the number of bits sent to the decoding device 200 can be reduced, and encoding efficiency can be improved.

[0659] For example, in inter-frame prediction modes where motion information of the target block is not directly encoded, skipping modes and / or merging modes may exist. Here, the motion information of each of the neighboring units in the encoding device 100 and decoding device 200 that can be used to indicate reconstruction will be used as the identifier and / or index of the unit's motion information for the target unit.

[0660] 2) Merge mode Merging is a scheme used to derive motion information for a target block. The term "merging" can mean merging the motion of multiple blocks. Merging can also mean that the motion information of one block is applied to other blocks. In other words, a merging pattern can be a mode for deriving the motion information of a target block from the motion information of neighboring blocks.

[0661] When using the merging mode, the encoding device 100 can use motion information from spatial candidates and / or temporal candidates to predict motion information of the target block. Spatial candidates may include reconstructed spatially adjacent blocks that are spatially adjacent to the target block. Spatially adjacent blocks may include left-side adjacent blocks and top-side adjacent blocks. Temporal candidates may include col blocks. The terms "spatial candidate" and "spatial merging candidate" are used interchangeably and have the same meaning. The terms "temporal candidate" and "temporal merging candidate" are used interchangeably and have the same meaning.

[0662] The encoding device 100 can obtain a prediction block through prediction. The encoding device 100 can encode a residual block, which is the difference between the target block and the prediction block.

[0663] 2-1) Creation of a list of merge candidates When using the merging mode, each of the encoding device 100 and the decoding device 200 can create a merging candidate list using motion information from spatial candidates and / or motion information from temporal candidates. The motion information may include 1) a motion vector, 2) a reference frame index, and 3) a reference direction. The reference direction can be unidirectional or bidirectional. The reference direction may represent an inter-frame prediction indicator.

[0664] The merge candidate list can include merge candidates. Merge candidates can be motion information. In other words, the merge candidate list can be a list that stores multiple pieces of motion information.

[0665] The merged candidate can be the motion information of multiple temporal and / or spatial candidates. In other words, the merged candidate list can include the motion information of temporal and / or spatial candidates, etc.

[0666] Furthermore, the merge candidate list may include new merge candidates generated by combining merge candidates that already exist in the merge candidate list. In other words, the merge candidate list may include new motion information generated by combining multiple motion information items that previously existed in the merge candidate list.

[0667] In addition, the merge candidate list may include history-based merge candidates. History-based merge candidates may be motion information of blocks that were encoded and / or decoded before the target block.

[0668] In addition, the list of merge candidates may include merge candidates based on the average of two merge candidates.

[0669] Merging candidates can be specific patterns for deriving inter-frame prediction information. Merging candidates can also be information indicating specific patterns for deriving inter-frame prediction information. Inter-frame prediction information for a target block can be derived based on the specific patterns indicated by the merging candidates. Furthermore, the specific patterns can include the process of deriving a series of inter-frame prediction information. Such specific patterns can be inter-frame prediction information deriving patterns or motion information deriving patterns.

[0670] Inter-frame prediction information for the target block can be derived based on the pattern indicated by the merge candidate selected in the merge candidate list via the merge index.

[0671] For example, the motion information export mode in the merge candidate list can be at least one of the following modes: 1) motion information export mode for sub-block units and 2) affine motion information export mode.

[0672] In addition, the list of merged candidates may include motion information for the zero vector. The zero vector may also be referred to as a "zero merged candidate".

[0673] In other words, the multiple motion information in the merged candidate list can be at least one of the following: 1) motion information of spatial candidates, 2) motion information of temporal candidates, 3) motion information generated by combining multiple motion information that previously existed in the merged candidate list, and 4) zero vector.

[0674] Motion information may include 1) motion vectors, 2) reference frame indexes, and 3) reference directions. The reference direction can also be referred to as an "inter-frame prediction indicator." The reference direction can be unidirectional or bidirectional. A unidirectional reference direction can indicate L0 prediction or L1 prediction.

[0675] A list of merge candidates can be created before performing predictions in merge mode.

[0676] The number of merge candidates in the merge candidate list can be predefined. Each of the encoding device 100 and the decoding device 200 can add merge candidates to the merge candidate list according to a predefined scheme and a predefined priority, so that the merge candidate list has a predefined number of merge candidates. The merge candidate list of the encoding device 100 and the merge candidate list of the decoding device 200 can be made identical to each other using a predefined scheme and a predefined priority.

[0677] Merging can be applied based on either CU or PU. When merging is performed based on CU or PU, the encoding device 100 can send a bitstream including predefined information to the decoding device 200. For example, the predefined information may include 1) information indicating whether merging is to be performed on each block partition, and 2) information about the blocks to be merged among the blocks that are spatial candidates and / or temporal candidates for the target block.

[0678] 2-2) Search for a motion vector using the list of merge candidates Encoding device 100 can determine merge candidates to be used for encoding a target block. For example, encoding device 100 can perform prediction on the target block using merge candidates from the merge candidate list and can generate residual blocks for the merge candidates. Encoding device 100 can encode the target block using the merge candidate that generates the minimum cost in encoding both the prediction and the residual blocks.

[0679] In addition, the encoding device 100 can determine whether to use a merge mode to encode the target block.

[0680] 2-3) Transmission of inter prediction information Encoding device 100 can generate a bitstream including inter-frame prediction information required for inter-frame prediction. Encoding device 100 can generate entropy-coded inter-frame prediction information by performing entropy coding on the inter-frame prediction information, and can send the bitstream including the entropy-coded inter-frame prediction information to decoding device 200. The entropy-coded inter-frame prediction information can be sent by encoding device 100 to decoding device 200 via a bitstream signal. Decoding device 200 can extract the entropy-coded inter-frame prediction information from the bitstream, and can obtain inter-frame prediction information by applying entropy decoding to the entropy-coded inter-frame prediction information.

[0681] The decoding device 200 can use the inter-frame prediction information of the bitstream to perform inter-frame prediction on the target block.

[0682] Inter-frame prediction information may include 1) mode information indicating whether a merging mode is used, 2) merging index, and 3) correction information.

[0683] In addition, inter-frame prediction information may include residual signals.

[0684] Decoding device 200 can obtain the merge index from the bitstream only when the mode information indicates that the merge mode is used.

[0685] Pattern information can be merge flags. The unit of pattern information can be a block. Information about a block can include pattern information, and the pattern information can indicate whether a merge pattern is applied to the block.

[0686] The merge index indicates which merge candidate from the merge candidate list will be used to predict the target block. Alternatively, the merge index may indicate which block from the spatially or temporally adjacent neighboring blocks will be merged with the target block.

[0687] Encoding device 100 can select the merge candidate with the highest encoding performance from the merge candidate list, and can set the value of the merge index to indicate the selected merge candidate.

[0688] The correction information can be information used to correct motion vectors. Encoding device 100 can generate the correction information. Decoding device 200 can correct the motion vectors of the merging candidates selected by the merging index based on the correction information.

[0689] The correction information may include at least one of information indicating whether correction will be performed, correction direction information, and correction size information. A prediction mode that corrects motion vectors based on correction information transmitted by a signal may be referred to as a "merging mode with motion vector difference".

[0690] 2-4) Inter prediction in merge mode using inter prediction information Decoding device 200 can perform prediction on target block using a merge candidate indicated by a merge index from among the merge candidates included in the merge candidate list.

[0691] The motion vector of the target block can be specified by the motion vector of the merge candidate indicated by the merge index, the reference screen index, and the reference direction.

[0692] 3) Skip mode Skip mode can be a mode that applies spatial or temporal motion information to the target block without alteration. Furthermore, skip mode can be a mode that does not use the residual signal. In other words, when using skip mode, the reconstructed block can be identical to the predicted block.

[0693] The difference between merge mode and skip mode lies in whether or not residual signals are sent or used. In other words, skip mode is similar to merge mode except that residual signals are not sent or used.

[0694] When using skip mode, encoding device 100 can transmit information about blocks whose motion information will be used as motion information for target blocks via a bitstream to decoding device 200. Encoding device 100 can generate entropy-encoded information by performing entropy encoding on this information, and can transmit the entropy-encoded information as a signal to decoding device 200 via a bitstream. Decoding device 200 can extract the entropy-encoded information from the bitstream, and can obtain information by applying entropy decoding to the entropy-encoded information.

[0695] Furthermore, when using the skip mode, the encoding device 100 may not send other syntax information (such as MVD) to the decoding device 200. For example, when using the skip mode, the encoding device 100 may not send the syntax elements associated with at least one of MVD, code block flag, and transform coefficient level to the decoding device 200.

[0696] 3-1) Creation of a list of merge candidates The merge candidate list can also be used in skip mode. In other words, the merge candidate list can be used in both merge mode and skip mode. In this respect, the merge candidate list can also be referred to as the "skip candidate list" or the "merge / skip candidate list".

[0697] Alternatively, the skip mode may use an additional candidate list that differs from the candidate list used in the merge mode. In this case, in the following description, the merge candidate list and merge candidate may be replaced by the skip candidate list and the skip candidate, respectively.

[0698] A list of merged candidates can be created before performing predictions in skip mode.

[0699] 3-2) Search for a motion vector using the list of merge candidates Encoding device 100 can determine merge candidates to be used for encoding the target block. For example, encoding device 100 can perform prediction on the target block using merge candidates from the merge candidate list. Encoding device 100 can encode the target block using the merge candidate that generates the minimum cost in the prediction.

[0700] In addition, the encoding device 100 can determine whether to use a skip mode to encode the target block.

[0701] 3-3) Transmission of inter prediction information Encoding device 100 can generate a bitstream that includes inter-frame prediction information required for inter-frame prediction. Decoding device 200 can use the inter-frame prediction information of the bitstream to perform inter-frame prediction on the target block.

[0702] Inter-frame prediction information may include 1) mode information indicating whether a skip mode is used and 2) a skip index.

[0703] Skipping indexes is the same as merging indexes as described above.

[0704] When using skip mode, the target block can be encoded without using the residual signal. Inter-frame prediction information may not include the residual signal. Alternatively, the bitstream may not include the residual signal.

[0705] Decoding device 200 may obtain the skip index from the bitstream only when the mode information indicates that a skip mode is used. As mentioned above, the merge index and the skip index may be the same. Decoding device 200 may obtain the skip index from the bitstream only when the mode information indicates that either a merge mode or a skip mode is used.

[0706] Skip index indicates which of the merge candidates included in the merge candidate list will be used to predict the target block.

[0707] 3-4) Inter prediction in skip mode using inter prediction information Decoding device 200 can perform prediction on the target block using a merge candidate indicated by a skip index from among the merge candidates included in the merge candidate list.

[0708] The motion vector of the target block can be specified by the motion vector of the merge candidate indicated by the skip index, the reference screen index, and the reference direction.

[0709] 4) Current picture reference mode The current frame reference mode can represent a prediction mode that uses the previously reconstructed area in the target frame to which the target block belongs.

[0710] Motion vectors can be used to specify previously reconstructed areas. The reference frame index of the target block can be used to determine whether the target block has been encoded in the current frame reference mode.

[0711] A flag or index indicating whether a target block is encoded in the current screen reference mode can be sent by the encoding device 100 to the decoding device 200. Alternatively, whether a target block is encoded in the current screen reference mode can be inferred from the target block's reference screen index.

[0712] When a target block is encoded in the current frame reference mode, the current frame can exist in a fixed position or any position in the reference frame list for the target block.

[0713] For example, the fixed position could be the position where the reference screen index value is 0 or the last position.

[0714] When the target image exists at any position in the list of reference images, an additional reference image index indicating such an arbitrary position can be sent by the encoding device 100 to the decoding device 200 by a signal.

[0715] 5) Sub-block merge mode Sub-block merging mode can be a mode that derives motion information from sub-blocks of the CU.

[0716] When applying the sub-block merging mode, the motion information of col-sub-blocks of the target sub-block in the reference image (i.e., based on the temporal merging candidate of the sub-block) and / or affine control point motion vector merging candidates can be used to generate a list of sub-block merging candidates.

[0717] 6) Triangular partition mode In the triangular partitioning mode, the target block can be partitioned diagonally, and sub-target blocks generated through partitioning can be produced. For each sub-target block, motion information of the corresponding sub-target block can be exported, and the exported motion information can be used to derive the prediction samples of each sub-target block. The prediction samples of the target block can be derived by weighted summing of the prediction samples of the sub-target blocks generated through partitioning.

[0718] 7) Combined inter-intra prediction mode The combined inter-frame-intra-frame prediction mode can be a mode that uses a weighted sum of prediction samples generated via inter-frame prediction and prediction samples generated via intra-frame prediction to derive prediction samples for the target block.

[0719] In the above mode, the decoding device 200 can autonomously correct the exported motion information. For example, the decoding device 200 can search for motion information with the minimum sum of absolute differences (SAD) in a specific region based on a reference block indicated by the exported motion information, and can export the found motion information as corrected motion information.

[0720] In the above mode, the decoding device 200 can use optical flow to compensate for the prediction samples derived via inter-frame prediction.

[0721] In the AMVP mode, merge mode, skip mode, etc. described above, the index information of the list can be used to specify the motion information among multiple motion information in the list that will be used to predict the target block.

[0722] To improve coding efficiency, the encoding device 100 may use only the index of the element in the signal transmission list that generates the minimum cost in inter-frame prediction of the target block. The encoding device 100 may encode this index and may signal the encoded index.

[0723] Therefore, the encoding device 100 and the decoding device 200 must be able to derive the lists described above (i.e., the candidate list for predicted motion vectors and the candidate list for merging) using the same scheme and based on the same data. Here, the same data may include reconstructed frames and reconstructed blocks. Furthermore, in order to specify elements using indices, the order of elements in the lists must be fixed.

[0724] Figure 10 Spatial candidates according to an embodiment are shown.

[0725] exist Figure 10 The image shows the locations of the spatial candidates.

[0726] The large block in the center of the graph represents the target block. The five smaller blocks represent spatial candidates.

[0727] The coordinates of the target block can be (xP, yP), and the size of the target block can be represented by (nPSW, nPSH).

[0728] Spatial candidate A0 can be a block adjacent to the lower left corner of the target block. A0 can be a block that occupies the pixel located at coordinates (xP-1, yP+nPSH).

[0729] Spatial candidate A1 can be the block to the left of the target block. A1 can be the bottommost block among the blocks to the left of the target block. Alternatively, A1 can be the block to the top of A0. A1 can be the block occupying the pixel located at coordinates (xP-1, yP+nPSH-1).

[0730] Spatial candidate B0 can be the block adjacent to the top right corner of the target block. B0 can be a block that occupies the pixel located at coordinates (xP+nPSW, yP-1).

[0731] Spatial candidate B1 can be the block that is top-adjacent to the target block. B1 can be the rightmost block among the blocks that are top-adjacent to the target block. Alternatively, B1 can be the block that is left-adjacent to B0. B1 can be the block that occupies the pixel located at coordinates (xP+nPSW-1, yP-1).

[0732] Spatial candidate B2 can be a block adjacent to the top-left corner of the target block. B2 can be a block that occupies the pixel located at coordinates (xP-1, yP-1).

[0733] Determination of the availability of spatial and temporal candidates In order to include spatial or temporal motion information in the list, it must be determined whether the spatial or temporal motion information is available.

[0734] In the following text, candidate blocks may include spatial candidates and temporal candidates.

[0735] For example, the determination can be performed by sequentially applying steps 1) through 4).

[0736] Step 1) When the PU including the candidate block is outside the boundary of the screen, the availability of the candidate block can be set to "false". The statement "availability is set to false" can have the same meaning as "set to unavailable".

[0737] Step 2) When the PU including the candidate block is outside the boundary of the stripe, the availability of the candidate block can be set to "false". When the target block and the candidate block are in different stripes, the availability of the candidate block can be set to "false".

[0738] Step 3) When the PU including the candidate block is outside the boundary of the parallel block, the availability of the candidate block can be set to "false". When the target block and the candidate block are in different parallel blocks, the availability of the candidate block can be set to "false".

[0739] Step 4) When the prediction mode of the PU including the candidate block is intra-frame prediction mode, the availability of the candidate block can be set to "false". When the PU including the candidate block does not use inter-frame prediction, the availability of the candidate block can be set to "false".

[0740] Figure 11 The order in which motion information of spatial candidates is added to the merging list is shown according to an embodiment.

[0741] like Figure 11 As shown, when multiple motion information entries from spatial candidates are added to the merge list, the order A1, B1, B0, A0, and B2 can be used. In other words, multiple motion information entries from available spatial candidates can be added to the merge list in the order A1, B1, B0, A0, and B2.

[0742] Method for deriving a merge list in merge and skip modes As described above, the maximum number of merge candidates in the merge list can be set. The maximum number can be indicated by "N". The set number can be sent from the encoding device 100 to the decoding device 200. The stripe header can include N. In other words, the maximum number of merge candidates in the merge list for the target block of the stripe can be set via the stripe header. For example, the value of N can essentially be 5.

[0743] Multiple motion information (i.e., merge candidates) can be added to the merge list in the order of steps 1) to 4).

[0744] Step 1) Within the space candidates, available space candidates can be added to the merge list. This can be done by... Figure 11 The order shown illustrates how multiple motion information entries for available space candidates are added to the merge list. Here, if the motion information of an available space candidate overlaps with other motion information already existing in the merge list, the motion information of the available space candidate may not be added to the merge list. The operation of checking whether the corresponding motion information overlaps with other motion information existing in the list can be simply referred to as "overlap check".

[0745] The maximum number of motion information entries that can be added is N.

[0746] Step 2) When the number of motion information entries in the merge list is less than N and time candidates are available, the motion information of the time candidates can be added to the merge list. However, if the available motion information of time candidates overlaps with other motion information already existing in the merge list, the available motion information of time candidates may not be added to the merge list.

[0747] Step 3) When the number of motion information entries in the merge list is less than N and the target strip type is "B", the combined motion information generated by combining bidirectional prediction (double prediction) can be added to the merge list.

[0748] The target strip can be a strip that includes the target block.

[0749] Combined motion information can be a combination of L0 motion information and L1 motion information. L0 motion information can be motion information that only refers to the L0 reference frame list. L1 motion information can be motion information that only refers to the L1 reference frame list.

[0750] The merged list may contain one or more L0 motion entries. Additionally, the merged list may contain one or more L1 motion entries.

[0751] Combined motion information may include one or more pieces of combined motion information. When generating combined motion information, the L0 motion information and L1 motion information that will be used in the step of generating combined motion information can be predefined from the one or more L0 motion information and the one or more L1 motion information. One or more pieces of combined motion information can be generated in a predefined order via bidirectional prediction using a pair of different motion information from a merge list. One of the different motion information in the pair can be L0 motion information, and the other of the different motion information in the pair can be L1 motion information.

[0752] For example, the combined motion information with the highest priority can be a combination of L0 motion information with a merge index of 0 and L1 motion information with a merge index of 1. When the motion information with a merge index of 0 is not L0 motion information, or when the motion information with a merge index of 1 is not L1 motion information, neither combined motion information is generated nor added. Next, the combined motion information with the next higher priority can be a combination of L0 motion information with a merge index of 1 and L1 motion information with a merge index of 0. Subsequent detailed combinations can conform to other combinations in the field of video encoding / decoding.

[0753] Here, when the combined motion information overlaps with other motion information already existing in the merge list, the combined motion information may not be added to the merge list.

[0754] Step 4) When the number of motion information entries in the merge list is less than N, the motion information of the zero vector can be added to the merge list.

[0755] Zero-vector motion information can be motion information where the motion vector is zero.

[0756] The number of zero-vector motion information entries can be one or more. The reference frame indices for one or more zero-vector motion information entries can be different from each other. For example, the reference frame index value for the first zero-vector motion information entry can be 0. The reference frame index value for the second zero-vector motion information entry can be 1.

[0757] The number of zero-vector motion information entries can be the same as the number of reference frames in the reference frame list.

[0758] The reference direction for zero-vector motion information can be bidirectional. Both motion vectors can be zero vectors. The number of zero-vector motion information entries can be the smaller of the number of reference frames in reference frame list L0 and the number of reference frames in reference frame list L1. Alternatively, when the number of reference frames in reference frame list L0 and the number of reference frames in reference frame list L1 are different from each other, a unidirectional reference direction can be used for reference frame indexing that can be applied to only a single reference frame list.

[0759] Encoding device 100 and / or decoding device 200 may subsequently add zero-vector motion information to the merge list while changing the reference screen index.

[0760] When zero-vector motion information overlaps with other motion information already existing in the merge list, the zero-vector motion information may not be added to the merge list.

[0761] The order of steps 1) to 4) above is merely exemplary and can be changed. Furthermore, some steps in the above steps may be omitted based on predefined conditions.

[0762] Method for deriving a list of prediction motion vector candidates in AMVP mode The maximum number of predicted motion vector candidates in the candidate list can be predefined. N can be used to indicate the predefined maximum number. For example, the predefined maximum number could be 2.

[0763] Multiple pieces of motion information (i.e., predicted motion vector candidates) can be added to the predicted motion vector candidate list in the order of steps 1) to 3).

[0764] Step 1) Available spatial candidates can be added to the list of predicted motion vector candidates. Spatial candidates may include a first spatial candidate and a second spatial candidate.

[0765] The first spatial candidate can be one of A0, A1, scaled A0, and scaled A1. The second spatial candidate can be one of B0, B1, B2, scaled B0, scaled B1, and scaled B2.

[0766] Multiple motion information entries from available spatial candidates can be added to the predicted motion vector candidate list in the order of first spatial candidate and second spatial candidate. In this case, if the motion information of an available spatial candidate overlaps with other motion information already existing in the predicted motion vector candidate list, the motion information of the available spatial candidate may not be added to the predicted motion vector candidate list. In other words, when the value of N is 2, if the motion information of the second spatial candidate is the same as the motion information of the first spatial candidate, then the motion information of the second spatial candidate may not be added to the predicted motion vector candidate list.

[0767] The maximum number of motion information entries that can be added is N.

[0768] Step 2) When the number of motion information entries in the predicted motion vector candidate list is less than N and a time candidate is available, the motion information of the time candidate can be added to the predicted motion vector candidate list. In this case, if the motion information of the available time candidate overlaps with other motion information already existing in the predicted motion vector candidate list, the motion information of the available time candidate may not be added to the predicted motion vector candidate list.

[0769] Step 3) When the number of motion information entries in the candidate list of predicted motion vectors is less than N, zero-vector motion information can be added to the candidate list of predicted motion vectors.

[0770] Zero-vector motion information may include one or more zero-vector motion information pieces. The reference frame indices of the one or more zero-vector motion information pieces may be different from each other.

[0771] Encoding device 100 and / or decoding device 200 can sequentially add multiple zero-vector motion information to the candidate list of predicted motion vectors while changing the reference frame index.

[0772] When zero-vector motion information overlaps with other motion information already existing in the candidate list of predicted motion vectors, the zero-vector motion information may not be added to the candidate list of predicted motion vectors.

[0773] The description of zero-vector motion information presented above, combined with the merged list, can also be applied to zero-vector motion information. Repeated descriptions will be omitted.

[0774] The order of steps 1) to 3) described above is merely exemplary and can be changed. Furthermore, some steps may be omitted based on predefined conditions.

[0775] Figure 12 The transformation and quantization process is shown based on the example.

[0776] like Figure 12 As shown, quantization levels can be generated by performing transformation and / or quantization processes on the residual signal.

[0777] The residual signal can be generated as the difference between the original block and the predicted block. Here, the predicted block can be a block generated via intra-frame prediction or inter-frame prediction.

[0778] The residual signal can be transformed into a signal in the frequency domain through a transformation process that is part of the quantization process.

[0779] Transform kernels used for transformations can include various DCT kernels, such as Discrete Cosine Transform (DCT) Type 2 (DCT-II) and Discrete Sine Transform (DST) kernels.

[0780] These transform kernels can perform separable or two-dimensional (2D) non-separable transforms on the residual signal. A separable transform can be a transform that indicates performing a one-dimensional (1D) transform on the residual signal in each of the horizontal and vertical directions.

[0781] In addition to DCT-II, the DCT and DST types adaptively used for 1D transformations may also include DCT-V, DCT-VIII, DST-I, and DST-VII, as shown in each of Tables 3 and 4 below.

[0782] Table 3

[0783] Table 4

[0784] As shown in Tables 3 and 4, transform sets can be used when deriving the DCT or DST type to be used for the transform. Each transform set can include multiple transform candidates. Each transform candidate can be a DCT type or a DST type.

[0785] Table 5 below shows examples of the transform sets that will be applied in the horizontal direction and the transform sets that will be applied in the vertical direction according to the intra-frame prediction mode.

[0786] Table 5

[0787] Table 5 shows the numbers of the vertical transform set and horizontal transform set that will be applied to the horizontal direction of the residual signal according to the intra-frame prediction mode of the target block.

[0788] like Figure 4 and Figure 5 As illustrated, the transform sets to be applied in the horizontal and vertical directions can be predefined based on the intra-prediction mode of the target block. Encoding device 100 can perform transforms and inverse transforms on the residual signal using transforms included in the transform set corresponding to the intra-prediction mode of the target block. Furthermore, decoding device 200 can perform inverse transforms on the residual signal using transforms included in the transform set corresponding to the intra-prediction mode of the target block.

[0789] In the transform and inverse transform, as illustrated in Tables 3, 4, and 5, the set of transforms to be applied to the residual signal can be determined and may not be transmitted by signal. Transform indication information can be transmitted from the encoding device 100 to the decoding device 200 by signal. The transform indication information may be information indicating which of the multiple transform candidates included in the set of transforms to be applied to the residual signal is used.

[0790] For example, when the target block size is 64×64, transform sets with three transforms can be configured according to the intra-frame prediction mode. The optimal transform method can be selected from a total of nine multi-transform methods generated by combinations of three transforms in the horizontal direction and three transforms in the vertical direction. With such an optimal transform method, the residual signal can be encoded and / or decoded, thus improving coding efficiency.

[0791] Here, information indicating which of the multiple transformations belonging to each transform set has been used for at least one of the vertical and horizontal transformations can be entropy encoded and / or entropy decoded. Here, truncated univariate binarization can be used to encode and / or decode such information.

[0792] As mentioned above, various transformation methods can be applied to residual signals generated via intra-frame prediction or inter-frame prediction.

[0793] The transformation may include at least one of a first transformation and a second transformation. Transform coefficients can be generated by performing a first transformation on the residual signal, and second transformation coefficients can be generated by performing a second transformation on the transform coefficients.

[0794] The first transformation can be referred to as the "primary transformation". Furthermore, the first transformation can also be referred to as the "Adaptive Multitransformation (AMT) scheme". As mentioned above, AMT can represent applying different transformations to various 1D directions (i.e., the vertical and horizontal directions).

[0795] A secondary transformation can be a transformation used to increase the energy concentration of the transformation coefficients generated by the first transformation. Similar to the first transformation, a secondary transformation can be a separable transformation or a non-separable transformation. Such a non-separable transformation can be a non-separable secondary transformation (NSST).

[0796] The first transformation can be performed using at least one of a predefined plurality of transformation methods. For example, the predefined plurality of transformation methods may include the Discrete Cosine Transform (DCT), the Discrete Sine Transform (DST), the Karhunen-Loeve Transform (KLT), etc.

[0797] Furthermore, depending on the kernel function defined for the Discrete Cosine Transform (DCT) or Discrete Sine Transform (DST), the first transform can be of various types.

[0798] For example, the transform type can be determined based on at least one of the following: 1) the prediction mode of the target block (e.g., one of intra-frame prediction and inter-frame prediction), 2) the size of the target block, 3) the shape of the target block, 4) the intra-frame prediction mode of the target block, 5) the components of the target block (e.g., one of luma component and chroma component), and 6) the partition type applied to the target block (e.g., one of quadtree, binary tree and ternary tree).

[0799] For example, based on the transform kernels presented in Table 6 below, the first transform may include transforms such as DCT-2, DCT-5, DCT-7, DST-7, DST-1, DST-8, and DCT-8. Table 6 below illustrates various transform types and transform kernel functions used for Multiple Transform Selection (MTS).

[0800] MTS can refer to the selection of a combination of one or more DCT and / or DST cores to transform the residual signal in the horizontal and / or vertical directions.

[0801] Table 6

[0802] In Table 6, i and j can be integer values ​​that are equal to or greater than 0 and less than or equal to N-1.

[0803] A secondary transformation can be performed on the transformation coefficients generated by performing the first transformation.

[0804] For example, in the first transformation, a transformation set can also be defined in the secondary transformation. The methods used to derive and / or determine the above transformation set can be applied not only to the first transformation but also to the secondary transformation.

[0805] The first and second transformations can be determined for a specific target.

[0806] For example, the first and second transformations can be applied to one or more signal components corresponding to the luma and chroma components. Whether to apply the first and / or second transformations can be determined based on at least one of the coding parameters for the target block and / or neighboring blocks. For example, whether to apply the first and / or second transformations can be determined based on the size and / or shape of the target block.

[0807] In the encoding device 100 and the decoding device 200, transformation information indicating the transformation method to be used for the target can be derived by utilizing specified information.

[0808] For example, the transformation information may include transformation indices that will be used for primary and / or secondary transformations. Alternatively, the transformation information may indicate that primary and / or secondary transformations are not used.

[0809] For example, when the target of the primary and secondary transforms is a target block, the transform method to be applied to the primary and / or secondary transforms, as indicated by the transform information, can be determined based on at least one of the encoding parameters for the target block and / or blocks adjacent to the target block.

[0810] Alternatively, the encoding device 100 may send transformation information indicating the transformation method for a specific target to the decoding device 200 via a signal.

[0811] For example, for a single CU, the decoding device 200 can derive transformation information such as whether a primary transformation is used, the index indicating the primary transformation, whether a secondary transformation is used, and the index indicating the secondary transformation. Alternatively, for a single CU, transformation information indicating the following can be transmitted via signals: whether a primary transformation is used, the index indicating the primary transformation, whether a secondary transformation is used, and the index indicating the secondary transformation.

[0812] Quantized transform coefficients (i.e., quantization levels) can be generated by quantizing the result produced by performing a first transform and / or a secondary transform, or by quantizing the residual signal.

[0813] Figure 13 This shows a diagonal scan based on an example.

[0814] Figure 14 The horizontal scan is shown based on the example.

[0815] Figure 15 The vertical scan is shown according to the example.

[0816] The quantized transform coefficients can be scanned via at least one of (top right) diagonal scan, vertical scan, and horizontal scan, based on at least one of intra-frame prediction mode, block size, and block shape. The block can be a transform unit (TU).

[0817] Each scan can be started at a specific start point and terminated at a specific end point.

[0818] For example, by using Figure 13 A diagonal scan is used to scan the coefficients of the block to transform the quantized transform coefficients into a 1D vector form. Alternatively, this can be used depending on the block size and / or intra-frame prediction mode. Figure 14 Horizontal scan or Figure 15 It uses vertical scanning instead of diagonal scanning.

[0819] A vertical scan can be an operation that scans 2D block-type coefficients in the column direction. A horizontal scan can be an operation that scans 2D block-type coefficients in the row direction.

[0820] In other words, the choice between diagonal, vertical, and horizontal scans can be determined based on the block size and / or inter-frame prediction mode.

[0821] like Figure 13 , Figure 14 and Figure 15 As shown, the quantized transform coefficients can be scanned along the diagonal, horizontal, or vertical direction.

[0822] The quantized transformation coefficients can be represented by block shapes. Each block can include multiple sub-blocks. Each sub-block can be defined based on either the minimum block size or the minimum block shape.

[0823] During scanning, the scanning order, based on the type or direction of the scan, can be applied first to the sub-blocks. Furthermore, the scanning order, based on the direction of the scan, can be applied to the quantized transform coefficients within each sub-block.

[0824] For example, such as Figure 13 , Figure 14 and Figure 15As shown, when the target block size is 8×8, quantized transform coefficients can be generated by a first transform, a second transform, and quantization of the residual signal of the target block. Therefore, one of the three types of scan sequences can be applied to four 4×4 sub-blocks, and the quantized transform coefficients can be scanned for each 4×4 sub-block according to the scan sequence.

[0825] Encoding device 100 can generate entropy-coded quantized transform coefficients by performing entropy coding on scanned quantized transform coefficients, and can generate a bit stream including the entropy-coded quantized transform coefficients.

[0826] The decoding device 200 can extract entropy-encoded quantized transform coefficients from the bitstream, and can generate quantized transform coefficients by performing entropy decoding on the entropy-encoded quantized transform coefficients. The quantized transform coefficients can be arranged in a 2D block format via inverse scanning. Here, as a method of inverse scanning, at least one of upper right diagonal scanning, vertical scanning, and horizontal scanning can be performed.

[0827] In the decoding device 200, inverse quantization can be performed on the quantized transform coefficients. A secondary inverse transform can be performed on the result generated by inverse quantization, depending on whether a secondary inverse transform is performed. Furthermore, a first inverse transform can be performed on the result generated by the secondary inverse transform, depending on whether a first inverse transform will be performed. The reconstructed residual signal can be generated by performing a first inverse transform on the result generated by the secondary inverse transform.

[0828] For luminance components reconstructed via intra-frame prediction or inter-frame prediction, an inverse mapping with dynamic range can be performed before loop filtering.

[0829] The dynamic range can be divided into 16 equal segments, and the mapping functions for the corresponding segments can be signaled. These mapping functions can be signaled at the stripe level or the parallel block group level.

[0830] An inverse mapping function can be derived from the mapping function to perform the inverse mapping.

[0831] Loop filtering, reference frame storage, and motion compensation can be performed in the inverse mapping region.

[0832] Predicted blocks generated via inter-frame prediction can be transformed to a mapped region using a mapping function, and the transformed predicted blocks can be used to generate reconstructed blocks. However, since intra-frame prediction is performed in the mapped region, predicted blocks generated via intra-frame prediction can be used to generate reconstructed blocks without requiring mapping and / or inverse mapping.

[0833] For example, when the target block is a residual block of the chrominance component, the residual block can be transformed into the inverse mapping region by scaling the chrominance component of the mapping region.

[0834] Scaling availability can be signaled at the stripe level or the parallel block group level.

[0835] For example, scaling can be applied only when the mapping is available for the luminance component and the partitions of the luminance and chrominance components follow the same tree structure.

[0836] Scaling can be performed based on the average value of the samples in the luminance prediction block corresponding to the chrominance prediction block. Here, when the target block uses inter-frame prediction, the luminance prediction block can represent the mapped luminance prediction block.

[0837] The scaling values ​​can be derived by using an index-referenced lookup table of the segment to which the average value of the sample values ​​of the brightness prediction block belongs.

[0838] The residual block can be transformed into the inverse mapping region by scaling the residual block using the final derived values. Subsequently, for blocks of the chroma components, reconstruction, intra-frame prediction, inter-frame prediction, loop filtering, and storage of reference frames can be performed in the inverse mapping region.

[0839] For example, information indicating whether the mapping and / or inverse mapping of the luminance and chrominance components is available can be sent via a sequence parameter set using signals.

[0840] A predicted block for the target block can be generated based on a block vector. The block vector indicates the displacement between the target block and a reference block. The reference block can be a block in the target image.

[0841] In this way, the prediction mode that generates prediction blocks by referencing the target image can be called the "intra-block copy (IBC) mode".

[0842] The IBC mode can be applied to CUs with specific dimensions. For example, the IBC mode can be applied to an M×N CU. Here, M and N can be less than or equal to 64.

[0843] IBC modes can include skip mode, merge mode, AMVP mode, etc. In skip mode or merge mode, a merge candidate list can be configured, and the merge index is signaled, allowing a single merge candidate to be specified from among the existing merge candidates in the merge candidate list. The block vector of the specified merge candidate can be used as the block vector of the target block.

[0844] In AMVP mode, the differential block vector can be signaled. Additionally, the prediction block vector can be derived from the target block's left and top neighboring blocks. Furthermore, the index of which neighboring block will be used can be signaled.

[0845] In IBC mode, the predicted block can be included in the target CTU or the left CTU, and can be limited to blocks within the previously reconstructed region. For example, the value of the block vector can be restricted such that the predicted block of the target block is located in a specific region. The specific region can be defined by three 64×64 blocks that are encoded and / or decoded before the 64×64 block including the target block. Restricting the value of the block vector in this way reduces memory consumption and device complexity caused by the implementation of IBC mode.

[0846] Figure 16 This is a configuration diagram of an encoding device according to an embodiment.

[0847] Encoding device 1600 can correspond to encoding device 100 described above.

[0848] Encoding device 1600 may include a processing unit 1610, a memory 1630, a user interface (UI) input device 1650, a UI output device 1660, and a storage 1640 that communicate with each other via a bus 1690. Encoding device 1600 may also include a communication unit 1620 connected to a network 1699.

[0849] The processing unit 1610 may be a central processing unit (CPU) or a semiconductor device for executing processing instructions stored in memory 1630 or storage 1640. The processing unit 1610 may be at least one hardware processor.

[0850] The processing unit 1610 can generate and process signals, data, or information input to, output from, or used in the encoding device 1600, and can perform checks, comparisons, determinations, etc., related to the signals, data, or information. In other words, in this embodiment, the processing unit 1610 can perform the generation and processing of data or information, as well as the checks, comparisons, and determinations related to the data or information.

[0851] The processing unit 1610 may include an inter-frame prediction unit 110, an intra-frame prediction unit 120, a switcher 115, a subtractor 125, a transform unit 130, a quantization unit 140, an entropy coding unit 150, an inverse quantization unit 160, an inverse transform unit 170, an adder 175, a filter unit 180, and a reference frame buffer 190.

[0852] At least some of the following components—inter-frame prediction unit 110, intra-frame prediction unit 120, switcher 115, subtractor 125, transform unit 130, quantization unit 140, entropy coding unit 150, inverse quantization unit 160, inverse transform unit 170, adder 175, filter unit 180, and reference frame buffer 190—may be program modules and capable of communicating with external devices or systems. These program modules may be included in the encoding device 1600 in the form of an operating system, application module, or other program modules.

[0853] The program modules can be physically stored in various types of known storage devices. Furthermore, at least some of the program modules can also be stored in a remote storage device capable of communicating with the encoding device 1600.

[0854] Program modules may include, but are not limited to, routines, subroutines, programs, objects, components, and data structures for performing functions or operations according to an embodiment or for implementing abstract data types according to an embodiment.

[0855] The program module can be implemented using instructions or code that run by at least one processor of the encoding device 1600.

[0856] The processing unit 1610 can execute instructions or codes in the inter-frame prediction unit 110, intra-frame prediction unit 120, switcher 115, subtractor 125, transform unit 130, quantization unit 140, entropy coding unit 150, dequantization unit 160, inverse transform unit 170, adder 175, filter unit 180 and reference frame buffer 190.

[0857] The storage unit may represent memory 1630 and / or storage 1640. Each of memory 1630 and storage 1640 may be any of a variety of volatile or non-volatile storage media. For example, memory 1630 may include at least one of read-only memory (ROM) 1631 and random access memory (RAM) 1632.

[0858] The storage unit can store data or information used for the operation of the encoding device 1600. In an embodiment, the data or information of the encoding device 1600 can be stored in the storage unit.

[0859] For example, storage units can store images, blocks, lists, motion information, inter-frame prediction information, bitstreams, etc.

[0860] The encoding device 1600 can be implemented in a computer system that includes a computer-readable storage medium.

[0861] The storage medium may store at least one module required for the operation of the encoding device 1600. The memory 1630 may store at least one module and may be configured such that the at least one module is operated by the processing unit 1610.

[0862] The communication unit 1620 can be used to perform functions related to communication of data or information with the encoding device 1600.

[0863] For example, communication unit 1620 can send a bit stream to decoding device 1700, which will be described later.

[0864] Figure 17 This is a configuration diagram of a decoding device according to an embodiment.

[0865] Decoding device 1700 can be used in conjunction with decoding device 200 as described above.

[0866] The decoding device 1700 may include a processing unit 1710, a memory 1730, a user interface (UI) input device 1750, a UI output device 1760, and a storage 1740 that communicate with each other via a bus 1790. The decoding device 1700 may also include a communication unit 1720 connected to a network 1799.

[0867] Processing unit 1710 may be a central processing unit (CPU) or semiconductor device for executing processing instructions stored in memory 1730 or storage 1740. Processing unit 1710 may be at least one hardware processor.

[0868] The processing unit 1710 can generate and process signals, data, or information input to, output from, or used in the decoding device 1700, and can perform checks, comparisons, determinations, etc., related to the signals, data, or information. In other words, in this embodiment, the processing unit 1710 can perform the generation and processing of data or information, as well as the checks, comparisons, and determinations related to the data or information.

[0869] The processing unit 1710 may include an entropy decoding unit 210, an inverse quantization unit 220, an inverse transform unit 230, an intra-frame prediction unit 240, an inter-frame prediction unit 250, a switcher 245, an adder 255, a filter unit 260, and a reference frame buffer 270.

[0870] At least some of the entropy decoding unit 210, inverse quantization unit 220, inverse transform unit 230, intra-frame prediction unit 240, inter-frame prediction unit 250, adder 255, switcher 245, filter unit 260, and reference frame buffer 270 of decoding device 200 may be program modules and are capable of communicating with external devices or systems. These program modules may be included in decoding device 1700 in the form of an operating system, application module, or other program modules.

[0871] The program modules can be physically stored in various types of known storage devices. Furthermore, at least some of the program modules can also be stored in a remote storage device capable of communicating with the decoding device 1700.

[0872] Program modules may include, but are not limited to, routines, subroutines, programs, objects, components, and data structures for performing functions or operations according to an embodiment or for implementing abstract data types according to an embodiment.

[0873] The program module can be implemented using instructions or code executed by at least one processor of the decoding device 1700.

[0874] The processing unit 1710 can run instructions or codes in the entropy decoding unit 210, the dequantization unit 220, the inverse transform unit 230, the intra-frame prediction unit 240, the inter-frame prediction unit 250, the switcher 245, the adder 255, the filter unit 260, and the reference frame buffer 270.

[0875] The storage unit may represent memory 1730 and / or storage 1740. Each of memory 1730 and storage 1740 may be any of a variety of volatile or non-volatile storage media. For example, memory 1730 may include at least one of ROM 1731 and RAM 1732.

[0876] The storage unit can store data or information used for the operation of the decoding device 1700. In an embodiment, the data or information of the decoding device 1700 can be stored in the storage unit.

[0877] For example, storage units can store images, blocks, lists, motion information, inter-frame prediction information, bitstreams, etc.

[0878] The decoding device 1700 can be implemented in a computer system that includes a computer-readable storage medium.

[0879] The storage medium may store at least one module required for the operation of the decoding device 1700. The memory 1730 may store at least one module and may be configured such that the at least one module is operated by the processing unit 1710.

[0880] The communication unit 1720 can be used to perform functions related to communication of data or information with the decoding device 1700.

[0881] For example, communication unit 1720 can receive bit streams from encoding device 1700.

[0882] In the following text, "processing unit" may refer to processing unit 1610 of encoding device 1600 and / or processing unit 1710 of decoding device 1700. For example, regarding prediction-related functions, the processing unit may represent switch 115 and / or switch 245. Regarding inter-frame prediction-related functions, the processing unit may represent inter-frame prediction unit 110, subtractor 125, and adder 175, and may also represent inter-frame prediction unit 250 and adder 255. Regarding intra-frame prediction-related functions, the processing unit may represent intra-frame prediction unit 120, subtractor 125, and adder 175, and may also represent intra-frame prediction unit 240 and adder 255. Regarding transform-related functions, the processing unit may represent transform unit 130 and inverse transform unit 170, and may also represent inverse transform unit 230. Regarding quantization-related functions, the processing unit may represent quantization unit 140 and inverse quantization unit 160, and may also indicate inverse quantization unit 220. Regarding functions related to entropy encoding and / or entropy decoding, the processing unit may represent entropy encoding unit 150 and / or entropy decoding unit 210. Regarding functions related to filtering, the processing unit may represent filter unit 180 and / or filter unit 260. Regarding functions related to reference frames, the processing unit may instruct reference frame buffer 190 and / or reference frame buffer 270.

[0883] Figure 18 This is a flowchart illustrating a target block prediction method and a bit stream generation method according to an embodiment.

[0884] The target block prediction method and bitstream generation method according to the embodiment can be executed by the encoding device 1600. This embodiment can be part of a target block coding method or a video coding method.

[0885] In step 1810, the processing unit 1610 can determine the prediction information to be applied to the encoding of the target block.

[0886] The prediction information may include information used for the predictions described above. For example, the prediction information may include inter-frame prediction information. For example, the prediction information may include intra-frame prediction information.

[0887] In step 1820, the processing unit 1610 may use information about the target block and determined prediction information to perform a prediction on the target block.

[0888] Predictive blocks can be generated by predicting target blocks.

[0889] It can generate residual blocks as the difference between the target block and the predicted block. Information about the target block can be generated by applying transformations and quantization to the residual blocks.

[0890] Information about the target block may include the transform coefficients and quantization coefficients for the target block. Information about the target block may also include prediction information.

[0891] In addition, a reconstruction block can be generated as the sum of the prediction block and the reconstruction residual block.

[0892] In step 1830, the processing unit 1610 can generate a bit stream.

[0893] The bitstream may include information about the target block. Furthermore, the bitstream may include the information described above in the embodiments. For example, the bitstream may include encoding parameters related to the target block and / or the attributes of the target block.

[0894] Information included in the bitstream can be generated in step 1830, or information included in the bitstream can be generated at least partially in steps 1810 and 1820.

[0895] The processing unit 1610 can store the generated bitstream in the memory 1640. Alternatively, the communication unit 1620 can send the bitstream to the decoding device 1700.

[0896] The bitstream may include encoded information about the target block. The processing unit 1610 can generate encoded information about the target block by performing entropy encoding on the information about the target block.

[0897] Figure 19 This is a flowchart of a target block prediction method using a bitstream according to an embodiment.

[0898] The target block prediction method using a bitstream, according to an embodiment, can be executed by the decoding device 1700. This embodiment can be part of a target block decoding method or a video decoding method.

[0899] In step 1910, communication unit 1720 can obtain the bit stream. Communication unit 1720 can receive the bit stream from encoding device 1600.

[0900] The bit stream may include information about the target block.

[0901] Information about the target block may include the transform coefficients and quantization coefficients for the target block. Information about the target block may also include prediction information.

[0902] Furthermore, the bitstream may include the information described above in the embodiments. For example, the bitstream may include encoding parameters related to the target block and / or the attributes of the target block.

[0903] Computer-readable storage media may include bit streams, and information about a target block included in the bit stream may be used to perform prediction and decoding of the target block.

[0904] The bitstream may include encoded information about the target block. The processing unit 1710 can generate information about the target block by performing entropy decoding on the encoded information about the target block.

[0905] The processing unit 1710 can store the obtained bit stream in the memory 1740.

[0906] In step 1920, the processing unit 1710 can determine the prediction information to be applied to the decoding of the target block.

[0907] The processing unit 1710 can use the method used in the above embodiments to determine the prediction information.

[0908] The processing unit 1710 can determine prediction information for the target block based on information related to the prediction method obtained from the bit stream.

[0909] Prediction information may include inter-frame prediction information. Prediction information may also include intra-frame prediction information.

[0910] In step 1930, the processing unit 1710 can use the information about the target block and the determined prediction information to perform a prediction on the target block.

[0911] In step 1930, a prediction block can be generated by performing a prediction on the target block using the prediction information.

[0912] In addition, a reconstruction block can be generated as the sum of the prediction block and the reconstruction residual block.

[0913] Prediction of chroma blocks To reduce the number of bits required for encoding / decoding in chroma block prediction methods, chroma prediction can be performed using cross-component prediction methods such as the following: Cross-component linear model (CCLM), Gradient linear model (GLM), Filter-based linear model (FLM), and Convolutional cross-component model (CCCM).

[0914] In this embodiment, information from the luminance block can be used to perform prediction of the chrominance block. This prediction reduces the number of bits required to predict the chrominance block and improves the efficiency of signal transmission / encoding / decoding within the chrominance block.

[0915] In the embodiments, "luma" can refer to "luminance". "chroma" can refer to "chroma".

[0916] Information in embodiments In an embodiment, block information may refer to at least one of information about neighboring (surrounding) blocks, information about a reference block, and information about the current block. In an embodiment, the current block may refer to the target block.

[0917] In addition, block information may include at least one of the encoding parameters.

[0918] The coding parameters may include at least one of multiple pieces of information used in inter-frame prediction, intra-frame prediction, transform, inverse transform, quantization, dequantization, entropy coding / decoding, and loop filtering.

[0919] Encoding parameters may also include block index, intra-prediction mode candidate list, intra-prediction candidate index, etc.

[0920] The block shape in the encoding parameters indicates which of the square and non-square shapes is the block shape.

[0921] The motion vector can be a motion vector for at least one of L0, L1, L2 and L3.

[0922] The motion vector difference can be the motion vector difference of at least one of L0, L1, L2 and L3.

[0923] The reference image index can be a reference image index for at least one of L0, L1, L2 and L3.

[0924] The precision of the motion vector representation can be expressed in units of n samples and 1 / n samples (such as integer samples, 1 / 2 samples, 1 / 4 samples, 1 / 8 samples, 1 / 16 samples, and 1 / 32 samples). n can be a positive integer. Furthermore, n can be determined based on at least one of the encoding parameters and candidate encoding parameters. Additionally, n can be preset by the encoding device 1600 and the decoding device 1700, and can be a value transmitted from the encoding device 1600 to the decoding device 1700 by a signal.

[0925] Matching methods Figure 20 This shows the template configuration in the example inter-frame prediction.

[0926] Figure 21 The template configuration in intra-frame prediction according to an embodiment is shown.

[0927] Figure 22 This shows the template configuration in the example bilateral matching.

[0928] In an embodiment, the matching technique may refer to a method of calculating error cost while adjusting the position of the template using templates defined between comparison targets, and may include template matching (TM), bilateral matching (BM), etc.

[0929] likeFigure 20 and Figure 21 As shown, template matching can use the surrounding pixels of the current block to configure the current template for the current block, and can use the pixels in the search area of ​​the reference image to configure the reference template that matches the current template.

[0930] like Figure 22 As shown, two-sided matching can use pixels from a reference image to configure the template. Here, at least one of the reconstructed pixels from the current image or the reference image can be used to configure the template.

[0931] Figure 23 This shows the first configuration of the template in the template matching example.

[0932] Figure 24 This shows the second configuration of the template in the template matching example.

[0933] Figure 25 This shows the first configuration of a template with alternating empty lines in the template matching example.

[0934] Figure 26 This shows a second configuration of a template with alternating empty lines in the template matching example.

[0935] Figure 27 This illustrates a second configuration of a template with alternating empty pixels in template matching based on the example.

[0936] Figure 28 This illustrates a second configuration of a template with alternating empty pixels in template matching based on the example.

[0937] exist Figures 23 to 28 In the middle, the large rectangle in the lower right part represents a block. This block can be the current block. Small rectangles indicate the pixels of the template. The area of ​​the small rectangle indicates the template. The area adjacent to the top of the block can be the upper (top) template. The area adjacent to the left of the block can be the left template. Small shaded rectangles indicate the pixels used to configure the template. Small blank rectangles indicate pixels not used to configure the template.

[0938] Figure 29 This shows the template configuration in the example bilateral matching.

[0939] Figure 30 This shows the configuration of a template with alternating empty lines in a bilateral match based on the example.

[0940] Figure 31 This shows the configuration of a template with alternating empty pixels in a bilateral matching example.

[0941] exist Figures 29 to 31In the diagram, large rectangles represent templates. Small rectangles indicate the pixels of the template. Small shaded rectangles indicate the pixels used to configure the template. Small blank rectangles indicate pixels not used to configure the template.

[0942] In this matching technology, such as Figure 25 , Figure 26 , Figure 27 , Figure 28 , Figure 30 and Figure 31 As shown, the template can be configured to have alternating empty lines or empty pixels.

[0943] Each template can be configured to have one or more empty pixels or one or more empty lines.

[0944] The number of pixels, the number of lines, and the shape of the template that constitute the corresponding template can be determined based on different encoding parameters.

[0945] For example, the shape of the template can be configured differently based on the partition shape of the current block or neighboring blocks, the size of the current block or neighboring blocks, and the statistical value of the size.

[0946] For example, a template can be configured with edges whose size is equal to the size of the edge that contacts the current block or a neighboring block.

[0947] For example, a template can be configured with edges whose size is smaller than the size of the edges that contact the current block or neighboring blocks.

[0948] For example, a template can be configured with edges whose size is larger than the size of the edges that contact the current block or neighboring blocks.

[0949] For example, a template can be configured using the maximum, minimum, and median values ​​of the edge dimensions of each of the current block and neighboring blocks as the dimensions of each edge.

[0950] For example, when the size of the current block or neighboring blocks is less than a threshold, the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0951] In an embodiment, reducing lines (or pixels) may mean removing some of a plurality of lines (or pixels). Alternatively, reducing lines (or pixels) may mean reducing the number of lines (pixels) by applying interpolation, sampling, and / or filtering to the plurality of lines (or pixels).

[0952] For example, when the size of the current block or neighboring blocks is less than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be reduced.

[0953] In an embodiment, adding a line (or pixel) may mean copying some of a plurality of lines (or pixels). Alternatively, adding a line (or pixel) may mean increasing the number of lines (pixels) by applying interpolation, sampling, and / or filtering to the plurality of lines (or pixels).

[0954] For example, when the size of the current block or neighboring blocks is greater than a threshold, the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0955] For example, when the size of the current block or neighboring blocks is greater than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be reduced.

[0956] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[0957] For example, the shape of the template can be configured differently based on the motion information of neighboring blocks or the statistical values ​​of motion information.

[0958] For example, when one of the statistical values ​​of the motion vectors of neighboring blocks is less than a threshold, the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0959] For example, when one of the statistical values ​​of the motion vectors of neighboring blocks is less than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0960] For example, when one of the statistical values ​​of the motion vectors of neighboring blocks is greater than a threshold, the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0961] For example, when one of the statistical values ​​of the motion vectors of neighboring blocks is greater than a threshold, the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0962] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[0963] Figure 32 This illustrates the template configuration when performing intra-frame prediction in inter-frame images, based on an example.

[0964] Figure 33 This shows the template configuration when performing inter-frame prediction on inter-frame images, based on an example.

[0965] The number of pixels and lines constituting the template, as well as the shape of the template, can be determined differently based on the prediction patterns of the current block or neighboring blocks.

[0966] For example, when performing inter-frame prediction on the current block, at least one of the blocks determined to use inter-frame prediction can be selected, and the selected blocks can be used to configure the template.

[0967] For example, when performing inter-frame prediction on the current block, at least one of the blocks determined to use intra-frame prediction can be selected, and the selected blocks can be used to configure the template.

[0968] For example, when performing intra-frame prediction on the current block (in an inter-frame image (inter-frame picture)), at least one of the blocks determined to use inter-frame prediction can be selected, and the selected blocks can be used to configure the template.

[0969] For example, when performing intra-frame prediction on the current block (in an inter-frame image (inter-frame picture)), at least one of the blocks determined to use intra-frame prediction can be selected, and the selected blocks can be used to configure the template.

[0970] The number of pixels and lines constituting the template, as well as the shape of the template, can be determined differently based on the position of pixels, the distance between pixels, and the partitioned areas.

[0971] For example, the shape of the template can be configured differently depending on its distance from the center of the search area.

[0972] For example, when the distance from the center of the search area to the location where the match is performed is less than a threshold, the template can be configured by reducing the number of lines or pixels, or the size of the template can be reduced.

[0973] For example, when the distance from the center of the search area to the location where the match is performed is less than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0974] For example, when the distance from the center of the search area to the location where the match is performed is greater than a threshold, the template can be configured by reducing the number of lines or pixels, or the size of the template can be reduced.

[0975] For example, when the distance from the center of the search area to the location where the match is performed is greater than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0976] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[0977] For example, the shape of the template can be configured differently based on the distance from the pixel to which the first match is performed.

[0978] For example, when the distance to the first pixel to be matched is less than a threshold, the template can be configured by reducing the number of lines or pixels, or the size of the template can be reduced.

[0979] For example, when the distance to the first pixel to be matched is less than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0980] For example, when the distance from the first pixel to be matched is greater than a threshold, the template can be configured by reducing the number of lines or pixels, or the size of the template can be reduced.

[0981] For example, when the distance from the first pixel to be matched is greater than a threshold, the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0982] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[0983] For example, when the search area is divided into multiple partitions, the shape of the template can be configured differently for each partition.

[0984] The number of pixels and lines constituting the corresponding template, as well as the shape of the template, can be determined differently based on the steps of performing the matching.

[0985] For example, when using template matching to find the nearest neighbor block with the lowest error cost, template matching can be performed only on specific locations within the search range that belong to the defined pattern, as in the case of pattern matching methods, to reduce complexity.

[0986] Here, the pattern matching method can be performed through the following steps.

[0987] [Step 1] First, sparse matching can be performed by using large patterns to perform matching in the search area, or by performing matching after setting the distance between the positions to be matched to a large value.

[0988] [Step 2] Next, based on the position with the lowest error cost derived at [Step 1], fine matching can be performed either by using small patterns or by performing matching after the distance between the positions to which matching is applied is set to a small value.

[0989] The matching region to be performed in [Step 2] can be determined based on the position derived at [Step 1]. The position derived at [Step 1] can be the position with the lowest error cost among the positions in the search region.

[0990] In [Step 1], the template can be configured by reducing lines or pixels, or the size of the template can be reduced.

[0991] In [Step 2], the template can be configured by adding lines or pixels, or the size of the template can be increased.

[0992] Furthermore, in [Step 1], the template can be configured by adding lines or pixels, or the size of the template can be increased, and in [Step 2], the template can be configured by reducing lines or pixels, or the size of the template can be decreased.

[0993] Matching based on matching techniques can be performed at at least one pixel location within the search range.

[0994] For example, a match can be performed on at least one of the pixel positions derived from the candidate list.

[0995] For example, a match can be performed on at least one of the pixel locations that exist within a specific distance from the current block.

[0996] For example, a match can be performed on at least one pixel location that exists at a specific distance from the current block.

[0997] For example, when the search range is partitioned, a match can be performed on at least one of the pixel positions within a specific partition region.

[0998] For example, a match can be performed on at least one of the pixel positions belonging to a specific pattern.

[0999] For example, a match can be performed on at least one of the pixel positions derived from a block (or candidate) having the same or similar encoding parameter values ​​as the current block.

[1000] For example, when the motion information of a pixel location is the same as or similar to the motion information of a block that includes that pixel location, matching can be performed at that pixel location.

[1001] For example, when the prediction information for a pixel location is the same as or similar to the prediction information for a block that includes that pixel location, a match can be performed at that pixel location. The prediction information can be inter-frame prediction information and / or intra-frame prediction information.

[1002] Matching can be performed at a pixel location when the predicted direction of the pixel location is the same as or similar to the predicted direction of the block that includes that pixel location.

[1003] Matching can be performed at a pixel location when the intra prediction mode of the pixel location is the same as or similar to the intra prediction mode of the block that includes that pixel location.

[1004] The processes described above for template matching and bilateral matching can be performed based on the statistical values ​​of the candidate encoding parameters. For example, the statistical values ​​can be, and may include, the statistical values ​​of the candidate encoding parameters.

[1005] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1006] Figure 34 This is a flowchart of the prediction method according to an embodiment.

[1007] Steps 3410, 3420, 3430, and 3440 can be performed by each of the encoding device 1600 and the decoding device 1700. Steps 3410, 3420, 3430, and 3440 can be performed by the processing unit 1610 of the encoding device 1600 and the processing unit 1710 of the decoding device 1700.

[1008] Determining the prediction information in steps 1810 and 1920 may include steps 3410, 3420 and 3430, which will be described later.

[1009] Performing the prediction in steps 1820 and 1930 may include step 3440, which will be described later.

[1010] Alternatively, the setup and deriving of prediction-related information described in steps 3410, 3420, 3430, and 3440 may be included in each of steps 1810 and 1920.

[1011] Alternatively, the execution of the predictions described in steps 3410, 3420, 3430 and 3440 may be included in each of steps 1820 and 1930.

[1012] In step 3410, a prediction method and / or prediction model for the chromaticity signal can be determined.

[1013] The prediction model can be a cross-component prediction model. In this embodiment, the prediction model can be abbreviated as "model".

[1014] In step 3420, the information used to derive the prediction model can be determined.

[1015] Information used to derive a predictive model can include a reference region and representative values.

[1016] In step 3430, prediction merging using the prediction model can be performed.

[1017] The prediction model can be a cross-component prediction model.

[1018] Predictive merging can be cross-component predictive merging.

[1019] In step 3440, a chromaticity prediction signal can be generated using a cross-component prediction model. The chromaticity prediction signal may include chromaticity prediction blocks.

[1020] Bitstreams may include encoded information.

[1021] Each of steps 3410, 3420, 3430, and 3440 can be performed based on the encoded information. The information generated in steps 3410, 3420, 3430, and 3440 can be generated based on the encoded information. In step 3410, a prediction method and / or prediction model for the chroma signal can be determined based on the encoded information. In step 3420, information for deriving the prediction model can be determined based on the encoded information. In step 3430, prediction merging using the prediction model can be performed based on the encoded information. In step 3440, a chroma prediction signal can be generated using the prediction model based on the encoded information.

[1022] Determination of a prediction method for chroma signals The prediction method for the chromaticity signal is determined in step 3410 below.

[1023] In an embodiment, the current chroma block can be the chroma block that is currently being encoded / decoded / transmitted / predicted. The current chroma block can refer to the current chroma prediction block.

[1024] When determining the prediction method for a chromaticity signal, a prediction signal (or prediction block) can be generated based on the correlation across color components.

[1025] When predicting chromaticity signals based on cross-color component relationships, the predicted signals can be derived based on the cross-component linear model (CCLM).

[1026] The CCLM-based chromaticity signal prediction method can use the sample values ​​of reconstructed chromaticity samples adjacent to the current chromaticity block and reconstructed luminance samples at positions corresponding to the reconstructed chromaticity samples to generate a chromaticity prediction block for chromaticity prediction.

[1027] In an embodiment, a sample may include 1) a pixel, 2) statistical values ​​of pixel values ​​of one or more pixels, and 3) encoded information derived from one or more pixels. The encoded information may include encoding parameters, and in an embodiment may include information for encoding / decoding / transmitting signals.

[1028] The following can be used to derive a linear regression model: 1) the correlation between the surrounding pixels of the luma block and the surrounding pixels of the chroma block, or 2) statistical values ​​such as the maximum and minimum values, as shown in [Equation 1] below, and the model coefficients a and b can be calculated. The values ​​of the model coefficients a and b, along with the pixels in the luma block, can be used to derive the current chroma prediction block.

[1029] [Equation 1] C'(i, j) = a L'(i, j) + b or C'(i, j) = a L(x, y) + b C(i, j) can represent the current chromaticity prediction block or a sample point within the current chromaticity prediction block. L'(i, j) can represent the reconstructed luminance block / sample point corresponding to the position of the current chromaticity prediction block.

[1030] In this embodiment, the model may refer to a linear regression model. The model coefficient values ​​may include the values ​​of model coefficients a and b.

[1031] However, L'(i, j) can refer to a luminance sample that has been adjusted to have the same size as the chroma block through subsampling, downsampling, etc. L(x,y) can represent a luminance sample that has a luminance block with a size different from the chroma block.

[1032] Therefore, the luminance sample position (x,y) and chrominance sample position (i,j) of the luminance sample L(x,y) corresponding to the chrominance sample C'(i,j) can be different from each other.

[1033] When deriving the model coefficients of [Equation 1], the model coefficients can be derived separately for the U signal and the V signal.

[1034] When deriving the model coefficients of [Equation 1], model coefficients applicable to both U and V signals can be derived.

[1035] Whether a model is applied can be determined individually for each of the U and V signals. Information indicating whether a model is applied can be encoded / decoded / transmitted as a signal.

[1036] To derive the values ​​of model coefficients a and b, information about one or more of the following can be used: 1) the surrounding samples of the current chromaticity prediction block, 2) the surrounding samples of the luminance block corresponding to the position of the current chromaticity prediction block, and 3) the reference line index of the corresponding luminance block.

[1037] A reference line can refer to multiple reference lines, as shown above. Figure 8 As described, a reference line index can indicate a reference thread among multiple reference lines used for encoding / decoding / signaling a block. Reference lines can also refer to adjacent or neighboring sample lines of a block.

[1038] When determining the reference lines used to derive the model coefficients, N or more lines can be used as reference lines. Here, N can be a positive number of 0 or greater. N can be a preset value of the encoding device 1600 and / or the decoding device 1700, and can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1039] In an embodiment, the location of the neighboring blocks or samples to be used can be determined based on the chroma format.

[1040] For example, in YUV 4:4:4 or YUV 4:2:2 formats, a reference line for a luma block, indicated by a reference line index, and a reference line for the corresponding chroma block can be used. Here, corresponding reference lines can refer to reference lines with positions relative to the blocks. Alternatively, corresponding reference lines can refer to reference lines with the same number, such as... Figure 8 As shown.

[1041] For example, in the YUV 4:2:0 format, the index of the reference line of the chroma block (=IntraChromaRefLineIdx) can be exported, as shown in Equation 2 below.

[1042] [Equation 2] IntraChromaRefLineIdx = floor(IntraLumaRefLineIdx / 2) IntraLumaRefLineIdx can be the index of the adjacent sample line of the chroma block.

[1043] By utilizing only specific samples from the luminance reference line reconstructed based on a subsampling scheme, the number of luminance samples and the number of chrominance samples can be adjusted to be equal. In other words, the luminance sample corresponding to each chrominance sample can be determined. Here, the specific sample can be an even-numbered (or odd-numbered) pixel in the reference line.

[1044] In addition, luminance samples at specific sample locations can be identified as samples corresponding to chromaticity samples.

[1045] For example, when the size of the luma block is 2M×2N and the size of the chroma block is M×N, the luma block located at (M+N) X1 / 8, (M+N) X2 / 8, (M+N) X3 / 8 and (M+N) The sample point at position X4 / 8 can be identified as the luminance sample point corresponding to the chrominance sample point.

[1046] Here, Xi can be a specific constant used to determine the location of the sample points. i can be an integer of 1 or greater. The number of specific constants can be increased as the number of sample points increases.

[1047] For example, as illustrated above, when using 4 samples, the constants to X1 through X4 can be arbitrarily determined, as in the cases of (1,3,5,7) and (2,4,6,8). Alternatively, the constants to X1 through X4 can be determined differently based on the encoding information of neighboring blocks.

[1048] A luminance sample corresponding to a chrominance sample can be derived using N luminance samples. N can be a specific positive integer.

[1049] Figure 35 This illustrates downsampling based on a deep neural network, as shown in the example.

[1050] For example, downsampling can be used to determine the luminance sample corresponding to the chrominance sample.

[1051] Downsampling can be performed based on various filters, such as those in [Equation 3], [Equation 4] and [Equation 5] below.

[1052] [Equation 3]

[1053] [Equation 4]

[1054] [Equation 5]

[1055] It can be based on Figure 35 The deep neural network (DNN) shown uses filtering to perform downsampling.

[1056] It can calculate the statistical value of one or more of N luminance samples, and use at least one of the statistical values ​​to derive a representative luminance sample. The representative luminance sample can be used to replace the luminance sample.

[1057] In an embodiment, statistical values ​​may include the average, maximum, minimum, and median of the sample points.

[1058] For example, when the luminance samples corresponding to the adjacent sample c(0,-1) of the current chroma block are L(-1,-2), L(0,-2), L(1,-2), L(-1,-1), L(0,-1), and L(1,-1), the corresponding luminance samples can be used to extract candidate values ​​for deriving representative values. Representative values ​​may include average, maximum, minimum, median, etc.

[1059] Candidate values ​​used to derive representative values ​​can refer to statistical values. Representative values ​​can refer to the sample values ​​of the luminance samples corresponding to the chrominance samples used to derive the model coefficients.

[1060] For example, the reference line for chromaticity samples can be derived using the following [Equation 6].

[1061] [Equation 6] c(0,-1) = (L(-1, -2) + 2) L(0, -2) + L(1, -2) + L(-1, -1) + 2 L(0,-1) + L(1, -1) + 4 )>>3.

[1062] ">>" represents the right shift operator.

[1063] The luminance sample corresponding to the chrominance sample can be changed. Furthermore, the equation used to derive the chrominance prediction sample can be modified. Information about the luminance sample corresponding to the adjacent sample of the current chrominance block, or information about the equation used to derive the adjacent sample of the current chrominance block, can be transmitted from the encoding device 1600 to the decoding device 1700 via signals.

[1064] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal. Alternatively, the encoding information of surrounding samples or blocks can be used to derive the sample positions described in the embodiments.

[1065] Generation of a chroma prediction block When the values ​​of model coefficients a and b are derived, the current chromaticity prediction block can be generated using [Equation 1], etc.

[1066] As shown in [Equations 7], [Equations 8], and [Equations 9], the values ​​of a and b can be derived using at least one of N representative values ​​(e.g., a first representative value, a second representative value, a third representative value, and a fourth representative value). N can be a positive integer. For example, N can be 2 or 4.

[1067] [Equation 7] a = (CMAX - CMIN) / (LMAX - LMIN) [Equation 8] b = CMIN - a LMIN [Equation 9] a = Log2(CMAX - CMIN) - Log2(LMAX - LMIN)>>k The first representative value (=LMAX) refers to the x-axis value of the maximum value. The second representative value (=CMAX) refers to the y-axis value of the maximum value. The third representative value (=LMIN) refers to the x-axis value of the minimum value. The fourth representative value (=CMIN) refers to the y-axis value of the minimum value.

[1068] For example, to derive b in [Equation 8], the values ​​of LMAX and CMAX can be used.

[1069] In [Equation 7], division is used, and therefore it may be difficult to implement using integer units. Therefore, division can be replaced by multiplication and shift operations (=>>).

[1070] In [Equation 9], the value of a can be derived by using a subtraction operation that uses Log2, while avoiding the division operation in [Equation 7].

[1071] Since the value is derived from Log2, the value of a can be normalized by using the value of k in a shift operation (=>>).

[1072] Derivation of chroma signals / samples / blocks using multiple models In the embodiments, in CCLM-based chromaticity signal prediction, multiple models can be used to derive chromaticity signals / samples / blocks.

[1073] Multiple models can be derived from brightness samples at different locations.

[1074] For example, a model can be derived from sample points with even-numbered locations on the reference line. Conversely, another model can be derived from sample points with odd-numbered locations on the reference line.

[1075] For example, sample points located above the brightness block and sample points located to the left of the brightness block can be separated from each other, and different models can be derived separately using the separated sample points.

[1076] Derivation of multiple models based on statistical values of luma samples Multiple models can be derived based on the statistical values ​​of brightness samples.

[1077] For example, the average value of the brightness samples can be used as a threshold. Samples with values ​​less than or equal to the threshold and samples with values ​​exceeding the threshold can be separated from each other. Different models can be derived from the separated samples.

[1078] For example, the sample values ​​of brightness points can be sorted in a specific order. This specific order could be close to the average value. From the sorted brightness samples, the first N samples and the last S samples can be selected. Different models can be derived independently using the first N samples and the last S samples.

[1079] Here, N and S can be preset values ​​by the encoding device 1600 and the decoding device 1700, and can also be values ​​sent by the encoding device 1600 to the decoding device 1700 via signals.

[1080] Multiple models can be derived based on defined thresholds.

[1081] For example, a specific value can be used as a threshold. Samples with values ​​less than or equal to the threshold and samples with values ​​exceeding the threshold can be separated from each other. Different models can be derived from the separated samples.

[1082] The threshold can be a preset value of the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1083] In this embodiment, the range of sample values ​​can be from the minimum to the maximum value of the sample values. Alternatively, the range of sample values ​​can be the range between the minimum and the maximum value of the sample values.

[1084] For example, the range of sample values ​​can be divided into N groups. Different models can be derived for each group by utilizing the sample values ​​of the divided groups. In other words, models for each group can be derived based on the sample values ​​of the divided groups.

[1085] Here, N can be a value preset by the encoding device 1600 and the decoding device 1700, or it can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1086] Figure 36 This shows the range of sample values ​​divided into N groups based on the example.

[1087] For example, such as Figure 36 As shown, the sample points can be divided into groups of cells.

[1088] When sample points are divided into groups, for each group, the statistical values ​​of the sample points in that group can be used to derive a model for that group. Different models for each group can be derived based on the statistical values ​​of the sample points in that group.

[1089] For example, the average value of the samples in each group can be derived.

[1090] New sample groups can be generated based on the average value of the sample points in a group. Different models can be derived for each sample group.

[1091] Figure 37 The range of sample values ​​divided into two groups is shown based on the example.

[1092] It can export the average value of the samples in each group.

[1093] In an embodiment, the average sample value of each group can represent the average of the samples in the corresponding group. The total average sample value can represent the average of all samples. The average sample value can include the average of the sample values ​​of one or more groups and the total average sample value.

[1094] For each group, samples whose values ​​are less than or equal to the average of the sample values ​​and samples whose values ​​are greater than the average of the sample values ​​can be divided into sample groups.

[1095] For example, such as Figure 37 As shown, sample points can be configured into group 1 (sample points whose values ​​are less than or equal to the average value) and group 2 (sample points whose values ​​are greater than the average value) based on the average value of the sample points.

[1096] Different models can be exported for each group of sample points.

[1097] Figure 38 The model is shown as a new group derived from the example for the average values ​​of the sample points.

[1098] The range between the averages of two sample values ​​can be defined as a new group. In other words, a new group can include sample values ​​whose values ​​fall within the average of the two sample values. Optionally, a new group can include sample values ​​whose values ​​are greater than the average of one sample value and less than or equal to the average of the next sample value.

[1099] In other words, the average of the sample values ​​from the previous group can be used to determine the new group.

[1100] Here, as Figure 38 As shown, the average value of the total sample points can also be used as the average value of the sample points on which the groups are divided.

[1101] Different models can be derived for each of the newly created groups.

[1102] Figure 39 This shows a new group model derived from the average values ​​of sample points for the group, based on the example.

[1103] The range between the average values ​​of sample points in two groups can be defined as a new group. In other words, a new group may include samples whose values ​​fall within the average values ​​of the sample points in the two existing groups. Optionally, a new group may include samples whose values ​​are greater than the average value of one existing group and less than or equal to the average value of the next existing group.

[1104] In other words, the average of the sample values ​​from the previous group can be used to determine the new group.

[1105] Here, as Figure 39 As shown, the average value of the total sample points may not be used as the average value of the sample points on which the groups are divided.

[1106] Different models can be derived for each of the newly created groups.

[1107] Exporting multiple models When exporting multiple models, models exported from previously encoded / decoded blocks can be used.

[1108] In an embodiment, the term "model derived from a block" may refer to "model derived for a block" or "model derived from a block", and the terms "model from a block," "model for a block," and "model of a block" may be used interchangeably.

[1109] In an embodiment, the derivation of a model may refer to the computation of the model.

[1110] For example, for each block, the model derived from that block can be stored. The model derived from each encoded / decoded block can also be stored. When encoding / decoding the current block, the model of another stored block can be loaded and used in the current block.

[1111] In an embodiment, “a specific model was used for the block” may mean that a specific model was used as the model for that block.

[1112] Here, the model can be stored for each of all blocks. Optionally, the model of a block can only be stored if the chroma mode based on the cross-component model is used as the chroma mode of the block.

[1113] When determining a reference block for deriving the model of the current block, at least one block can be selected from either adjacent or non-adjacent blocks of the current block, and the selected block can be used as the reference block. Here, the storage of model coefficients in the embodiments can be applied to the derivation of the model.

[1114] Furthermore, when determining the reference block for exporting the model, at least one block can be selected from the blocks in the reference screen, and the selected block can be used as the reference block. Here, the storage of model coefficients in the embodiment can be applied to the export of the model.

[1115] When exporting multiple models, you can use models exported from blocks with dimensions different from the current block.

[1116] For example, the position of the reference region or reference points can be changed according to the size of the block, and different models can be configured based on the position of the reference region or reference points. Here, a model generated from a large-sized block can be loaded for a small-sized block, and the model can be used in the small-sized block. On the other hand, a model derived from a small-sized block can be used in a large-sized block.

[1117] In an embodiment, the reference region may include reconstructed samples, lines, and / or blocks for model parameter derivation and inheritance.

[1118] When exporting multiple models, a model exported from a block at another location within a partition can be used by that block.

[1119] For example, when the position of a reference region or reference sample point changes depending on the position of the block within a partitioned block, and different models are configured based on the position of the reference region or reference sample point, a model derived from a block at another location within the partitioned block can be used for the current block.

[1120] When exporting multiple models, you can export a new model by modifying the coefficient values ​​of the previously exported model.

[1121] Here, at least one of the coefficients, biases, and nonlinear terms constituting a new model can be derived by modifying the coefficients, biases, or nonlinear terms of the previously derived model.

[1122] For example, for the model "C'(i,j)=a" in the above [Equation 1] The model coefficients a and bias value b of L'(i,j)+b” can be used to generate a new model by modifying the values ​​of a or b.

[1123] Here, when the values ​​derived by modifying a and b are a' and b' respectively, a' and b' can be derived, as shown in [Equations 10] to [Equations 13] below.

[1124] [Equation 10] a' = a + v [Equation 11] a' = a - v [Equation 12] a' = a v [Equation 13] a' = a / v In other words, a' or b' can be derived by adding a specific value v to a or b. Alternatively, a' or b' can be derived by subtracting a specific value v from a or b. Alternatively, a' or b' can be derived by multiplying a specific value v by a or b. Alternatively, a' or b' can be derived by dividing a or b by a specific value v.

[1125] Here, a and b can be configured in polynomial form to derive an equation for a' or b', as shown in [Equation 14] below.

[1126] [Equation 14] a' = a t + s The coefficients (such as v, t, and s) used to derive a' or b' can be preset values ​​by the encoding device 1600 and / or the decoding device 1700, and can be values ​​sent from the encoding device 1600 to the decoding device 1700 by signals.

[1127] For a' and b', the range of values ​​for a' and b' can be set by W, X, Y, and Z, as shown in [Equations 15] and [Equations 16] below.

[1128] [Equation 15] W <a'<X [Equation 16] Y <b'<Z Here, each of W, X, Y and Z can be a value preset by the encoding device 1600 and / or the decoding device 1700, and can be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1129] When exporting multiple models, a weighted sum of the coefficient values ​​of previously exported models can be used to export a new model.

[1130] Here, at least one of the coefficients, biases, and nonlinear terms constituting the new model can be derived using a weighted sum of the coefficients, biases, or nonlinear terms of the previously derived model.

[1131] For example, the following [Model 1-1] and [Model 1-2] can represent the previously derived models.

[1132] [Model 1-1] C'(i, j) = a L'(i, j) + b [Model 1-2] C'(i, j) = c L'(i, j) + d Here, [Model 1-3] can represent the newly generated models using [Model 1-1] and [Model 1-2].

[1133] [Models 1-3] C'(i, j) = e L'(i, j) + f Here, the weighted sum of the coefficients in [Model 1-1] and [Model 1-2], and the weighted sum of [Model 1-1] and [Model 1-2], can be used to generate each of the coefficients e and deviations f in [Model 1-3].

[1134] The following [Equation 17] can represent the derivation of the coefficient e in [Model 1-3].

[1135] [Equation 17] e=w a + (1 - w) c The following [Equation 18] can represent the derivation of the deviation f in [Model 1-3].

[1136] [Equation 18] f=w b+(1-w) d In [Equations 17] and [Equations 18], w can represent the weight of the weighted sum.

[1137] The weight w can be a value preset by the encoding device 1600 and / or the decoding device 1700, and can also be a value sent from the encoding device 1600 to the decoding device 1700 by a signal.

[1138] The method for exporting models described in the embodiments can be used not only to export multiple models, but also to export a single model for the current block.

[1139] For example, a new model may not be available in the current block using reference samples, and a model previously computed in at least one of the blocks in an adjacent block, neighboring block, or reference frame may be used as the model for the current block.

[1140] For example, it may not be possible to derive a new model using surrounding samples in the current block, and a model derived by weighting and summing the coefficient values ​​of a previously derived model can be used as the model for the current block.

[1141] For example, a new model can be derived from surrounding samples that may not be available in the current block, and a model configured by modifying the coefficient values ​​of a previously exported model can be used as the model for the current block.

[1142] In the embodiments, a method for deriving cross-component prediction parameters using reference samples in a block at a specific location has been described. In this method, model parameters from a block at a specific location can be inherited to the current block.

[1143] Inherited model parameters can be used for cross-component prediction (CCP) merging of the current block.

[1144] In CCP merging, default candidates can be generated by scaling model parameters.

[1145] In an embodiment, scaling of model parameters can refer to changing model parameters according to a specific process (such as the operations in the embodiment). Here, operations can include multiplication and division, and can include addition and subtraction. For example, scaling of model parameters can include multiplying the value of a derived model parameter by an offset, dividing the derived model parameter by an offset, adding the value of a derived model parameter to an offset, and subtracting the value of a derived model parameter from an offset.

[1146] The model generated by the method described in the embodiments can be added to the merge candidate list for merging.

[1147] Model information determination Figure 40 The example shows the luminance blocks and the corresponding luminance blocks to the chrominance blocks.

[1148] Figure 41 Showing a chroma block based on the example.

[1149] Figure 42 The example shows a chromaticity prediction block.

[1150] Figure 43 The example shows the luminance block and the luminance block corresponding to the chrominance block.

[1151] Figure 44 Showing a chroma block based on the example.

[1152] Figure 45 The example shows four chromaticity prediction blocks.

[1153] Figure 41 and Figure 44 The chroma blocks in the image represent the size and pixels of the chroma block used as the prediction target.

[1154] Figure 42 A chromaticity prediction block can represent the predicted value of a chromaticity block generated by predicting the chromaticity block.

[1155] In the prediction of chroma signals based on multiple models, information about the model used to generate the chroma block among the multiple models can be sent from the encoding device 1600 to the decoding device 1700 via a signal. Alternatively, information about the model that will be used to generate the chroma block can be derived using the encoding information of the block or surrounding pixels.

[1156] Information about the model may include information about the equations used to derive the chromaticity signal, as well as information about the coefficients and constants included in the equations.

[1157] For example, multiple models such as [Model 1-1] to [Model 1-4] can be derived.

[1158] [Model 2-1] C'(i, j) = a L'(I, j) + b [Model 2-2] C''(i, j) = c L'(I, j) + d [Model 2-3] C'''(i, j) = e L'(I, j) + f [Model 2-4] C''''(i, j) = g L'(I, j) + h Multiple models can be used to generate regions for chroma blocks separately (e.g., by...). Figure 49 The predicted sample points (= C', C'', C''' and C'''') or chromaticity prediction blocks (= BLK_C', BLK_C'', BLK_C''', BLK_C''') of the region defined by the thick line in the image.

[1159] Chromaticity prediction allows selection of chromaticity prediction blocks and models with the lowest error cost. Here, the selected model can be one used to generate chromaticity prediction blocks with the lowest error cost.

[1160] Information indicating the selected model can be transmitted from the encoding device 1600 to the decoding device 1700 via a signal.

[1161] Signaling of model information When generating chroma signals / blocks based on multiple models, information about the models used to generate the final chroma signal / block can be sent using signals. The final chroma signal / block can refer to the chroma signal / block ultimately generated through prediction. The final chroma signal / block can be one of multiple signals / blocks that is ultimately determined as the result of prediction for a chroma block.

[1162] like Figures 40 to 45 As shown, multiple models can be used to generate multiple chromaticity prediction blocks. Multiple models can be used to perform chromaticity prediction that generates multiple chromaticity prediction blocks.

[1163] The final model can be determined based on the error cost of multiple chroma prediction blocks. The final model can be the model ultimately determined for predicting chroma blocks. Information about the final model can be transmitted via signals.

[1164] The error cost of each model can be the prediction cost of the chromaticity prediction blocks generated by the model.

[1165] For example, the model with the lowest error cost among multiple models can be determined as the final model.

[1166] Information indicating the final model can be sent using signals.

[1167] Here, the error cost can be calculated separately for the U signal and the V signal. Alternatively, the error cost for U and the error cost for V can be added together to form a single error cost.

[1168] Model information derivation method When generating chromaticity signals / blocks based on multiple models, a final model for generating the final chromaticity block can be derived by performing chromaticity prediction, which uses the process of deriving the model from the reference region and the process of applying the model to the reference region.

[1169] For example, the regions in the reference region from which the model is derived to extract information about the final model can be separated from the regions in which the error cost is calculated by applying the derived model, and the separated regions can be used separately.

[1170] Figure 46 The diagram shows the regions in the luminance block according to the example and the regions in the luminance block that directly correspond to the chrominance block.

[1171] Figure 47 The area shown is within the chroma block according to the example.

[1172] Figure 48 This shows the area in the luminance block that directly corresponds to the chroma block, based on the example.

[1173] Figure 49 The area shown is within the chroma block according to the example.

[1174] Figure 50 The example shows the chromaticity prediction block and chromaticity template.

[1175] Region R1 can be a region within the reference region used to derive the model.

[1176] Region R2 can be the region within the reference region where the chromaticity prediction signal is generated by applying the derived model and the error cost of the chromaticity signal is calculated.

[1177] One or more models can be derived using luminance samples in region R1 and the corresponding chrominance samples.

[1178] For example, region R2 can indicate a sample point in the reference region that is adjacent to the luminance block.

[1179] For example, region R1 can indicate a sample point in the reference region that is adjacent to region R2.

[1180] For example, region R2 can indicate a sample point in the reference region that is adjacent to the block. For example, a sample point in the reference region that is adjacent to one of the left, upper left, and top edges of the luminance block can be included in region R2.

[1181] For example, region R1 may indicate a sample point in the reference region that is adjacent to region R2. For example, a sample point in the reference region that is adjacent to one of the left, upper left, and top of region R2 may be included in region R1.

[1182] For example, region R1 can indicate a sample point in the reference region that is 2 units away from the block. Region R2 can indicate a sample point in the reference region that is 1 unit away from the block. Here, the distance between a specific sample point and a specific block can be the distance between the specific sample point and the nearest sample point in the specific block. The nearest sample point can be the sample point in the block that is closest to the specific sample point. The distance between a specific sample point and the block (or the nearest sample point) can be the greater of the horizontal distance between the sample point and the block (or the nearest sample point) and the vertical distance between the sample point and the block (or the nearest sample point).

[1183] Figure 51 This is a flowchart based on the example chromaticity prediction method.

[1184] In step 5110, the model coefficient values ​​of CCML can be derived using the samples included in region R1 among the reconstructed surrounding reference samples of the luminance block (= L') directly corresponding to the chrominance block (= C).

[1185] Here, model coefficient values ​​for multiple models can be derived by separating the sample points used to derive the model coefficients. For example, the sample points can be separated into even-numbered sample points and odd-numbered sample points.

[1186] In step 5120, multiple models can be used to generate chromaticity prediction samples or chromaticity prediction templates.

[1187] The exported model can be used to generate chromaticity prediction samples (e.g., C', C'', C''', and C'''') among the surrounding reference samples of the reconstructed chromaticity block (=C).

[1188] Chromaticity prediction samp...

Claims

1. An image decoding method, comprising: Determine the prediction method for chroma blocks; Determine the information used to derive the cross-component prediction model; and The cross-component prediction model is used to generate chromaticity prediction blocks.

2. The image decoding method as described in claim 1, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model that uses gradient and positional information.

3. The image decoding method as described in claim 1, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using multiple downsampling filters.

4. The image decoding method as described in claim 1, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using unsampled luminance samples.

5. The image decoding method as described in claim 1, wherein, Information from previously decoded blocks is used to configure a reference region for the cross-component prediction model.

6. The image decoding method as described in claim 1, wherein, The chromaticity prediction block is generated based on cross-component prediction merging using the cross-component prediction model.

7. The image decoding method as described in claim 1, wherein, The chromaticity prediction block is generated using a prediction block generated by a method other than the cross-component prediction model.

8. An image encoding method, comprising: Determine the prediction method for chroma blocks; Determine the information used to derive the cross-component prediction model; and The cross-component prediction model is used to generate chromaticity prediction blocks.

9. The image encoding method as described in claim 8, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model that uses gradient and positional information.

10. The image encoding method as described in claim 8, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using multiple downsampling filters.

11. The image encoding method as described in claim 8, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using unsampled luminance samples.

12. The image encoding method as described in claim 8, wherein, Information from previously encoded blocks is used to configure a reference region for the cross-component prediction model.

13. The image encoding method as described in claim 8, wherein, The chromaticity prediction block is generated based on cross-component prediction merging using the cross-component prediction model.

14. The image encoding method as described in claim 8, wherein, The chromaticity prediction block is generated using a prediction block generated by a method other than the cross-component prediction model.

15. A computer-readable storage medium for storing a bitstream for image decoding, wherein: The bitstream includes encoded information. The prediction method for chroma blocks is determined based on the encoded information. The information used to derive the cross-component prediction model was determined, and The chromaticity prediction block is generated using the cross-component prediction model.

16. The computer-readable storage medium of claim 15, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model that uses gradient and positional information.

17. The computer-readable storage medium of claim 15, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using multiple downsampling filters.

18. The computer-readable storage medium of claim 15, wherein, The chromaticity prediction block is generated based on a convolutional cross-component model using unsampled luminance samples.

19. The computer-readable storage medium of claim 15, wherein, Information from previously decoded blocks is used to configure a reference region for the cross-component prediction model.

20. The computer-readable storage medium of claim 15, wherein, The chromaticity prediction block is generated based on cross-component prediction merging using the cross-component prediction model.

Images