Method for storing and transmitting coded video bit stream

Abstract

To provide a technique for improving encoding efficiency by performing block division suitable for image encoding and decoding.SOLUTION: To provide a technique for improving encoding efficiency by performing block division suitable for image encoding and decoding. An affine inheritance merge candidate derivation unit configured to derive an affine inheritance merge candidate that inherits an affine model of a block spatially neighboring an encoding target block; and an affine fixed merge candidate derivation unit configured to derive an affine fixed merge candidate in which motion information of an affine control point is fixed, wherein the affine fixed merge candidate fixes a motion vector of each affine control point to (0,0).SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an image encoding and decoding technique that divides an image into blocks and performs prediction. [Background technology]

[0002] In image encoding and decoding, the image to be processed is divided into blocks, which are groups of a predetermined number of pixels. Divide into appropriate blocks and process in blocks. Encoding efficiency is improved by appropriately setting inter-frame prediction and frame prediction. do.

[0003] In video coding and decoding, inter-prediction is used, which predicts from pictures that have already been coded or decoded. In Patent Document 1, affine transformation is used in inter-prediction. The technology to be applied is described. In moving images, objects undergo transformations such as enlargement, reduction, and rotation. It is not uncommon for this to happen, and by applying the technology of Patent Document 1, efficient coding This becomes possible. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 9-172644 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the technology of Patent Document 1 involves image conversion, which places a heavy processing load on the system. In view of the above-mentioned problems, the present invention provides a low-load, efficient encoding technique. . [Means for solving the problem]

[0006] The embodiments described below may relate to the following aspects, for example. The first aspect is a storage method for storing on a recording medium a bitstream generated according to a video encoding method that performs affine transformation on a coding block basis, the video encoding method comprising: an affine inheritance merge candidate derivation step for deriving an affine inheritance merge candidate that inherits an affine model possessed by a block that is spatially adjacent to the block to be encoded; an affine construction merge candidate derivation step for deriving an affine construction merge candidate from multiple motion information possessed by blocks that are spatially or temporally adjacent to the block to be encoded; and an affine-fixed merge candidate derivation step for deriving an affine-fixed merge candidate in which motion information of affine control points is fixed, wherein the affine construction merge candidate derivation step fixes the reference image index to be compared to 0 and determines whether or not the affine transformation is possible.

[0007] A transmission method for transmitting a bitstream generated according to a video coding method that performs affine transformation on a coding block-by-block basis, the video coding method comprising: an affine inheritance merge candidate derivation step for deriving an affine inheritance merge candidate that inherits an affine model possessed by a block that is spatially adjacent to the block to be coded; an affine construction merge candidate derivation step for deriving an affine construction merge candidate from multiple motion information possessed by blocks that are spatially or temporally adjacent to the block to be coded; and an affine-fixed merge candidate derivation step for deriving an affine-fixed merge candidate in which motion information of affine control points is fixed, wherein the affine construction merge candidate derivation step fixes the reference image index to be compared to 0 and determines whether or not the affine transformation is possible.

[0008] It should be noted that these descriptions are merely examples. The scope of the present application and the present invention will not be limited by these descriptions. Furthermore, in this specification, the term "the present invention" The following should not be construed as limiting the scope of the present invention or application, but are used for illustrative purposes. It should be understood that: [Effects of the Invention]

[0009] According to the present invention, highly efficient image encoding and decoding processing can be realized with a low load. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram of an image encoding device according to an embodiment of the present invention; [Figure 2] 1 is a block diagram of an image decoding device according to an embodiment of the present invention. [Figure 3] 10 is a flowchart illustrating an operation of dividing a tree block. [Figure 4] FIG. 10 is a diagram illustrating how an input image is divided into tree blocks. [Figure 5] FIG. 1 is a diagram illustrating a z-scan. [Figure 6A] FIG. 10 is a diagram showing the divided shapes of blocks. [Figure 6B] FIG. 10 is a diagram showing the divided shapes of blocks. [Figure 6C] FIG. 10 is a diagram showing the divided shapes of blocks. [Figure 6D] FIG. 10 is a diagram showing the divided shapes of blocks. [Figure 6E] FIG. 10 is a diagram showing the divided shapes of blocks. [Figure 7] 10 is a flowchart illustrating an operation of dividing a block into four parts. [Figure 8] 10 is a flowchart illustrating an operation of dividing a block into two or three parts. [Figure 9] This is a syntax for expressing the shape of block division. [Figure 10A] FIG. 10 is a diagram illustrating intra prediction. [Figure 10B] FIG. 10 is a diagram illustrating intra prediction. [Figure 11] FIG. 10 is a diagram illustrating a reference block for inter prediction. [Figure 12] This is a syntax for expressing a coding block prediction mode. [Figure 13] FIG. 10 is a diagram showing correspondence between syntax elements and modes related to inter prediction. [Figure 14] FIG. 10 is a diagram for explaining affine transformation motion compensation of two control points. [Figure 15] FIG. 10 is a diagram for explaining affine transformation motion compensation of three control points. [Figure 16] FIG. 2 is a block diagram showing a detailed configuration of the inter prediction unit 102 in FIG. [Figure 17] FIG. 17 is a block diagram showing a detailed configuration of a normal predicted motion vector mode derivation unit 301 in FIG. [Figure 18] FIG. 17 is a block diagram showing a detailed configuration of a normal merge mode derivation unit 302 in FIG. 16. [Figure 19] 17 is a flowchart for explaining a normal predicted motion vector mode derivation process of the normal predicted motion vector mode derivation unit 301 of FIG. 16. [Figure 20] 10 is a flowchart showing a processing procedure of a normal predicted motion vector mode derivation process. [Figure 21] 10 is a flowchart illustrating a processing procedure for normal merge mode derivation processing. [Figure 22] FIG. 3 is a block diagram showing a detailed configuration of an inter prediction unit 203 in FIG. 2. [Figure 23] FIG. 23 is a block diagram showing a detailed configuration of a normal predicted motion vector mode derivation unit 401 in FIG. 22. [Figure 24] FIG. 23 is a block diagram showing a detailed configuration of a normal merge mode derivation unit 402 in FIG. 22. [Figure 25] 23 is a flowchart for explaining a normal predicted motion vector mode derivation process of the normal predicted motion vector mode derivation unit 401 of FIG. 22. [Figure 26] FIG. 10 is a diagram illustrating a procedure for initializing and updating a history motion vector predictor candidate list. [Figure 27] 10 is a flowchart of a procedure for checking identical elements in the procedure for initializing and updating a history motion vector predictor candidate list. [Figure 28] 10 is a flowchart of an element shifting process procedure in the history motion vector predictor candidate list initialization / update process procedure. [Figure 29] 10 is a flowchart illustrating a procedure for deriving a historical motion vector predictor candidate. [Figure 30] 10 is a flowchart illustrating a history merge candidate derivation process procedure. [Figure 31A] FIG. 10 is a diagram illustrating an example of a history motion vector predictor candidate list update process. [Figure 31B] FIG. 10 is a diagram illustrating an example of a history motion vector predictor candidate list update process. [Figure 31C] FIG. 10 is a diagram illustrating an example of a history motion vector predictor candidate list update process. [Figure 32] FIG. 10 is a diagram illustrating motion compensation prediction in the case of L0 prediction in which the L0 reference picture (RefL0Pic) is located at a time earlier than the current picture (CurPic). [Figure 33] FIG. 10 is a diagram illustrating motion compensation prediction in the case where L0 prediction is performed and the reference picture for L0 prediction is located at a later time than the current picture. [Figure 34] FIG. 10 is a diagram illustrating the prediction direction of motion compensation prediction in a bi-predictive case where the reference picture for L0 prediction is located at a time earlier than the current picture and the reference picture for L1 prediction is located at a time later than the current picture. [Figure 35] The prediction direction of the motion compensation prediction in the case of the bi-prediction video decoding method will be described below. [Figure 36] FIG. 10 is a diagram illustrating the prediction direction of motion compensation prediction in bi-prediction when the reference picture for L0 prediction and the reference picture for L1 prediction are located at a time later than the current picture. [Figure 37] 1 is a diagram illustrating an example of a hardware configuration of a coding / decoding device according to an embodiment of the present invention; [Figure 38]FIG. 3 is a block diagram of a sub-block predicted motion vector mode derivation unit 303 in the encoding device of the present application. [Figure 39] FIG. 4 is a block diagram of a sub-block predicted motion vector mode derivation unit 403 in the decoding device of the present application. [Figure 40] FIG. 3 is a block diagram of a sub-block merge mode derivation unit 304 in the encoding device of the present application. [Figure 41] FIG. 4 is a block diagram of a sub-block merge mode derivation unit 404 in the decoding device of the present application. [Figure 42] FIG. 10 is a diagram illustrating derivation of affine inheritance predicted motion vector candidates. [Figure 43] FIG. 10 is a diagram illustrating the derivation of affine-constructed predicted motion vector candidates. [Figure 44] FIG. 10 is a diagram illustrating affine inheritance merge candidate derivation. [Figure 45] FIG. 10 is a diagram illustrating affine construction merge candidate derivation. [Figure 46] 10 is a flowchart illustrating the derivation of affine inheritance predicted motion vector candidates. [Figure 47] 10 is a flowchart illustrating the derivation of affine constructed motion vector predictor candidates. [Figure 48] 10 is a flowchart of affine inheritance merge candidate derivation. [Figure 49] 10 is a flowchart of affine construction merge candidate derivation. [Figure 50] 10 is a flowchart illustrating the operation of a sub-block temporal merge candidate derivation unit 381. [Figure 51] 10 is a flowchart illustrating a process of deriving adjacent motion information of a block. [Figure 52] 10 is a flowchart illustrating a process of deriving a temporal motion vector. [Figure 53] 10 is a flowchart illustrating derivation of inter prediction information. [Figure 54] 10 is a flowchart illustrating a process of deriving sub-block motion information. [Figure 55]FIG. 1 is a diagram for explaining the temporal relationship between pictures. [Figure 56] 10 is a flowchart for explaining a process of deriving a temporal motion vector predictor candidate in a normal motion vector predictor mode derivation unit 301. [Figure 57] 10 is a flowchart for explaining the derivation process of ColPic in the derivation process of temporal motion vector predictor candidates in the normal motion vector predictor mode derivation unit 301. [Figure 58] 10 is a flowchart for explaining the process of deriving coding information of ColPic in the process of deriving a temporal motion vector predictor candidate in the normal motion vector predictor mode derivation unit 301. [Figure 59] 10 is a flowchart illustrating a process of deriving inter prediction information. [Figure 60] 10 is a flowchart showing the procedure of a process of deriving inter prediction information of a coding block when the inter prediction mode of the coding block colCb is bi-prediction (Pred_BI). [Figure 61] 10 is a flowchart illustrating a procedure for a motion vector scaling calculation process. [Figure 62] 10 is a flowchart illustrating a process of deriving temporal merge candidates. [Figure 63] FIG. 10 is a diagram showing a comparison target of motion information in deriving affine construction merge candidates. DETAILED DESCRIPTION OF THE INVENTION

[0011] The technologies and technical terms used in this embodiment will be defined below.

[0012] <Tree Block> In this embodiment, the image to be encoded / decoded is divided equally into a predetermined size. In Figure 4, the size of the treeblock is 128x128 pixels. However, the size of the tree block is not limited to this and can be any size. The processing target (the encoding target in the encoding process, the decoding target in the decoding process) may be set. The tree blocks of the tree are arranged in raster scan order, i.e., from left to right and from top to bottom. The order is as follows: Inside each tree block, further recursive division is possible. After recursively dividing the tree block, the block to be coded and decoded is called the coding block. In addition, tree blocks and coding blocks are collectively defined as blocks. By dividing the blocks appropriately, efficient coding becomes possible. The size of the code can be a fixed value previously agreed upon between the coding device and the decoding device, or The size of the treeblock determined by the encoding device may be transmitted to the decoding device. Here, the maximum size of the treeblock is set to 128x128 pixels, and the treeblock size is set to The minimum size of the coding block is 16x16 pixels. The maximum size of the coding block is 64x64. The minimum size of a pixel,coding block is set to 4x4 pixels.

[0013] <Prediction mode> In the coding block unit, prediction is performed from the processed image signal of the image to be processed. Intra prediction (MODE_INTRA), and inter prediction (MO) that predicts from the image signal of the processed image. DE_INTER). In the encoding process, a processed image is an image obtained by decoded a signal that has been encoded. It is used for codes, tree blocks, blocks, coding blocks, etc., and in the decoding process, is used for completed images, image signals, tree blocks, blocks, coding blocks, etc. . This mode distinguishes between intra prediction (MODE_INTRA) and inter prediction (MODE_INTER). The prediction mode (PredMode) is defined as intra prediction (MODE_INTRA ), or inter prediction (MODE_INTER) as a value.

[0014] <Inter prediction> Inter-prediction, which predicts from the image signal of a processed image, refers to multiple processed images. It can be used as a picture. To manage multiple reference pictures, L0 (reference link) Two types of reference lists are defined: L1 (reference list 1) and L2 (reference list 2), and each has a reference index. P slices can use L0 prediction (Pred_L0) For B slices, L0 prediction (Pred_L0), L1 prediction (Pred_L1), and bi-prediction (Pred_BI) are available. ) is available. L0 prediction (Pred_L0) refers to the reference picture managed by L0. Inter prediction is performed using the reference picture managed by L1, and L1 prediction (Pred_L1) is performed using the reference picture managed by L1. Bi-prediction (Pred_BI) is inter-prediction where both L0 and L1 predictions are performed. Inter prediction refers to one reference picture managed in L0 and one in L1. Information specifying L0 prediction, L1 prediction, or bi-prediction is defined as an inter prediction mode. In the following processing, the constants and variables with the subscript LX in the output are L0, L1 It is assumed that processing is performed for each

[0015] <Predictive motion vector mode> The predicted motion vector mode is an index for specifying the predicted motion vector, a differential motion vector, and a The inter prediction vector, inter prediction mode, and reference index are transmitted, and the index of the block to be processed is This is a mode for determining the motion vector prediction information of the target block. A processed block that belongs to the same image as the target block, or a block that belongs to the processed image and has the same position as the target block A motion vector predictor candidate derived from a block located at or near (neighboring) the predicted motion vector is used. The measured motion vector is derived from an index for identifying the measured motion vector.

[0016] <Merge mode> In merge mode, the target block is processed without transmitting differential motion vectors and reference indices. The processed blocks adjacent to the block to be processed or the blocks belonging to the processed image The inter prediction information of the block located at the same position as the block or in the vicinity (neighborhood) is used for processing. This is a mode for deriving inter prediction information for the current block.

[0017] The processed blocks adjacent to the target block and the inter-processing blocks The spatial merge candidate is defined as a block belonging to the processed image. Blocks located in the same location as the block or in its vicinity (neighborhood), and the index of that block Inter-prediction information derived from super-prediction information is defined as a temporal merge candidate. The complement is registered in the merge candidate list, and the prediction of the target block is determined by the merge index. Identify merge candidates to use in

[0018] <Adjacent Block> FIG. 11 shows the procedure for deriving inter prediction information in the predicted motion vector mode and merge mode. A0, A1, A2, B0, B1, B2, B3 is a processed block adjacent to the target block. T0 is a processed image. The block belongs to the same position as the processing target block in the processing target image or its vicinity ( It is a block located in the vicinity.

[0019] A1 and A2 are located on the left side of the coding block to be processed and are adjacent to the coding block to be processed. B1 and B3 are adjacent blocks located above the coding block to be processed. A0, B0, and B2 are the blocks adjacent to the target coding block. These are the blocks located at the bottom left, top right, and top left of the encryption block.

[0020] Details of how adjacent blocks are handled in predicted motion vector mode and merge mode The details will be described later.

[0021] <Affine transformation motion compensation> Affine transformation motion compensation divides a coding block into sub-blocks of a predetermined unit, and A motion vector is determined for each sub-block individually and motion compensation is performed. The motion vector of each sub-block is determined by the motion vector of the processed block adjacent to the current block, or is a block belonging to the processed image, which is at the same position as the target block or in its vicinity (neighborhood) based on one or more control points derived from inter prediction information of the block located at In this embodiment, the size of the sub-block is set to 4x4 pixels. The size is not limited to this, and the motion vector may be derived in units of pixels.

[0022] FIG. 14 shows an example of affine transformation motion compensation when there are two control points. A control point has two parameters, a horizontal component and a vertical component. The affine transformation in these two cases is called a four-parameter affine transformation. CP2 is the control point. FIG. 15 shows an example of affine transformation motion compensation when there are three control points. A control point has two parameters, a horizontal component and a vertical component. The affine transformation in the three cases is called a six-parameter affine transformation. CP2 and CP3 are control points.

[0023] Affine transformation motion compensation is available in both the predicted motion vector mode and the merge mode. It is also available in the prediction motion vector mode. The mode in which the affine transformation is performed is defined as the sub-block predicted motion vector mode, and the merge mode is defined as the affine transformation mode. A mode in which motion compensation is applied is defined as a sub-block merge mode.

[0024] <Inter prediction syntax> The syntax related to inter prediction will be described with reference to FIGS. The merge_flag in FIG. 12 indicates whether the current coding block is in merge mode or whether the predicted motion vector is merge_affine_flag is a flag that indicates whether to use the merge mode. This is a flag indicating whether or not to apply the sub-block merging mode to the current coding block. inter_affine_flag is the sub-block of the coding block to be processed in the predicted motion vector mode. cu_affine_type_flag A flag indicating whether to apply the affine motion vector prediction mode. is a flag for determining the number of control points in the sub-block predicted motion vector mode. is. Figure 13 shows the values ​​of each syntax element and the corresponding prediction method. ag=1,merge_affine_flag=0 corresponds to the normal merge mode. merge_flag=1,merge_affine_flag=1 is a merge mode that is not a subblock merge. merge_flag=0,inter_affine_flag=0 corresponds to the normal prediction mode. Normal predicted motion vector mode corresponds to the sub-block predicted motion vector mode. merge_flag=0,inter_affine_flag=1 corresponds to the sub-block predicted motion vector mode. merge_flag=0,inter_affine_flag If cu_affine_type_flag is set to 1, the number of control points is determined.

[0025] <poc> POC (Picture Order Count) is a variable associated with the picture being coded. The value increases by one depending on the picture output order. It can determine whether a picture is the same as another picture, determine the order of pictures in the output order, and For example, if two pictures have the same POC, we can calculate the distance between them. If two pictures have different POC values, they are considered to be the same picture. If it does, the picture with the smaller POC value is determined to be the picture to be output first. The difference between the POCs of two pictures indicates the distance between the pictures along the time axis.

[0026] (First embodiment) Regarding the image encoding device 100 and the image decoding device 200 according to the first embodiment of the present invention, I will explain.

[0027] FIG. 1 is a block diagram of an image coding device 100 according to a first embodiment. The image encoding device 100 includes a block division unit 101, an inter prediction unit 102, an intra prediction unit 103, and a A prediction unit 103, a decoded image memory 104, a prediction method determination unit 105, a residual generation unit 106, and an orthogonal A transform / quantization unit 107, a bit string encoding unit 108, an inverse quantization / inverse orthogonal transformation unit 109, and a decoding unit The image signal superimposing unit 110 and the coded information storage memory 111 are provided.

[0028] The block division unit 101 recursively divides the input image to generate coding blocks. The block division unit 101 divides the blocks to be divided into horizontal and vertical blocks. The four quadrants and the blocks to be divided can be arranged either horizontally or vertically. The block division unit 101 processes the generated coded blocks. The image signal of the processing target coding block is input to the inter prediction unit. 102, the intra prediction unit 103, and the residual generation unit 106. The unit 101 supplies information indicating the determined recursive division structure to the bit string encoding unit 108. The detailed operation of the lock dividing unit 101 will be described later.

[0029] The inter prediction unit 102 performs inter prediction on the coding block to be processed. The measurement unit 102 receives the inter-prediction information stored in the coded information storage memory 111 and the decoding information. A plurality of inter prediction information is calculated from the decoded image signal stored in the image memory 104. and deriving candidates for the inter prediction mode, selecting a suitable inter prediction mode from the derived candidates. and a predicted image signal according to the selected inter prediction mode. The inter prediction unit 102 supplies the data to the prediction method determination unit 105. The detailed configuration and operation of the inter prediction unit 102 will be described later. .

[0030] The intra prediction unit 103 performs intra prediction on the coding block to be processed. The measurement unit 103 measures the decoded image signal stored in the decoded image memory 104 as a reference pixel. and refers to the coding information stored in the coding information storage memory 111. A predicted image signal is generated by intra prediction based on the encoding information. The intra prediction unit 103 selects an appropriate intra prediction mode from among a plurality of intra prediction modes. Then, the selected intra prediction mode and a predicted image according to the selected intra prediction mode are generated. The image signal is supplied to a prediction method determination unit 105 . 10A and 10B show examples of intra prediction. For example, the intra prediction mode 5 0 generates an intra-predicted image by copying reference pixels vertically. Prediction mode 1 is a DC mode in which all pixel values ​​of the block to be processed are converted to the reference pixel values. Intra prediction mode 0 is a planar mode, and vertical This mode creates a two-dimensional intra-prediction image from reference pixels in the vertical and horizontal directions. 10B is an example of generating an intra-predicted image in the case of intra-prediction mode 40. The intra prediction unit 103 performs intra prediction on each pixel of the block to be processed in a direction indicated by the intra prediction mode. The intra prediction unit 103 copies the values ​​of the reference pixels in the intra prediction mode. If the position is not an integer, the reference pixel value is determined by interpolation from the reference pixel values ​​of the surrounding integer positions. Determine.

[0031] The decoded image memory 104 stores the decoded image generated by the decoded image signal superimposing unit 110 . The decoded image memory 104 transfers the stored decoded images to the inter prediction unit 102 and the intra prediction unit 103. The measurement unit 103 supplies the data.

[0032] The prediction method determination unit 105 determines the coding information for each of intra prediction and inter prediction. The evaluation is based on the amount of coding of the information and residual, the amount of distortion between the predicted image signal and the image signal to be processed, etc. In the case of intra prediction, the prediction method determination unit 10 5 is a bitstream encoding unit 1 which encodes intra prediction information such as an intra prediction mode as encoding information. In the case of the merge mode of inter prediction, the prediction method determination unit 105 supplies the merged sub-block merge mode (sub-block merge flag) The inter prediction information such as the inter frame coding information is supplied to the bitstream coding unit 108 as coding information. In the case of the predicted motion vector mode of the inter prediction, the prediction method determination unit 105 code, predicted motion vector index, L0, L1 reference index, differential motion vector information indicating whether it is a sub-block predicted motion vector mode (sub-block predicted motion vector) Inter prediction information such as a bit flag is supplied to the bitstream encoding unit 108 as encoding information. Furthermore, the prediction method determination unit 105 stores the determined coding information in the coding information storage memory 11. The prediction method determination unit 105 outputs the residual signal to the residual generation unit 106 and the prediction image signal to the decoded image 1. The signal is supplied to the signal superimposing unit 110.

[0033] The residual generation unit 106 generates a residual by subtracting a predicted image signal from an image signal to be processed. The generated signal is supplied to the orthogonal transform and quantization unit 107 .

[0034] The orthogonal transform and quantization unit 107 performs orthogonal transform and quantization on the residual in accordance with the quantization parameter. The generated residual is then input to a bitstream coding unit 10. 8 and the inverse quantization and inverse orthogonal transformation unit 109.

[0035] The bitstream coding unit 108 encodes the bitstream in units of sequences, pictures, slices, and coding blocks. In addition to the information, the prediction method determined by the prediction method determination unit 105 for each coding block is Specifically, the bit string coding unit 108 encodes the coding information according to the coding block. If the prediction mode is inter prediction (MODE_INTER), In this case, the bitstream encoding unit 108 sets a flag for determining whether or not the merge mode is selected, a sub-block merge merge flag, merge index if in merge mode, index if not in merge mode Sub-prediction mode, predicted motion vector index, information about differential motion vector, Encoding information (inter prediction information) such as block prediction motion vector flags is stored in the specified syntax. The first bitstream is generated by encoding it according to the bitstream syntax rules. If MODE_INTRA is intra prediction, coding information such as intra prediction mode (intra prediction The first bit is encoded according to a specified syntax (syntax rules for bit strings). The bit string coding unit 108 also generates a bit string by defining the orthogonally transformed and quantized residuals. The second bit string is generated by entropy encoding the bit string according to the syntax The encoding unit 108 multiplexes the first bit string and the second bit string according to a specified syntax. and outputs the bitstream.

[0036] The inverse quantization and inverse orthogonal transformation unit 109 performs the orthogonal transformation supplied from the orthogonal transformation and quantization unit 107. The quantized residual is inversely quantized and inversely orthogonally transformed to calculate the residual, and the calculated residual is used as the decoded image. The signal is supplied to the image signal superimposing unit 110 .

[0037] The decoded image signal superimposing unit 110 outputs a predicted image signal according to the determination by the prediction method determining unit 105. The residuals that have been inversely quantized and inversely orthogonally transformed by the inverse quantization and inverse orthogonal transformation unit 109 are superimposed to form a decoded image. The decoded image signal superimposing unit 110 generates a decoded image and stores it in the decoded image memory 104. The coded image is filtered to reduce block distortion and other distortions caused by coding. The decoded image may then be stored in the decoded image memory 104.

[0038] The encoding information storage memory 111 stores the prediction mode (integer) determined by the prediction method determination unit 105. In the case of inter prediction, the code The coding information stored in the coding information storage memory 111 includes the determined motion vectors, the reference list, Inter prediction information such as L0 and L1 reference indexes, historical predicted motion vector candidate list, etc. In the case of the merge mode of the inter prediction, the coding information storage memory 111 includes The encoded information to be stored includes the merge index, sub-block merge index, and the above information. Inter prediction information (sub-block merge flag) indicating whether the sub-block is in the multi-mode or not is included. In the case of the inter-prediction predicted motion vector mode, the coding information storage memory 111 The encoded information to be stored includes the above information as well as the inter prediction mode, predicted motion vector, etc. An index, a differential motion vector, and information indicating whether it is a sub-block predicted motion vector mode. It includes inter prediction information such as sub-block prediction motion vector flags. In the case of prediction, the coding information stored in the coding information storage memory 111 includes the determined input. It includes intra prediction information such as intra prediction mode.

[0039] FIG. 2 shows the configuration of an image decoding apparatus according to an embodiment of the present invention, which corresponds to the image encoding apparatus of FIG. The image decoding device according to the embodiment includes a bitstream decoding unit 201, a block block division unit 202, inter prediction unit 203, intra prediction unit 204, coding information storage memory 205, an inverse quantization and inverse orthogonal transformation unit 206, a decoded image signal superimposition unit 207, and a decoded image memory Equipped with 208 harpoons.

[0040] The decoding process of the image decoding device of FIG. 2 is performed by the decoding device provided inside the image coding device of FIG. Since it corresponds to the processing, the coding information storage memory 205 in FIG. The configurations of the conversion unit 206, the decoded image signal superimposing unit 207, and the decoded image memory 208 are shown in FIG. The image encoding device includes an encoding information storage memory 111, an inverse quantization and inverse orthogonal transformation unit 109, a decoder, and a The decoded image signal superimposing unit 110 and the decoded image memory 104 and their corresponding functions It has.

[0041] The bitstream supplied to the bitstream decoder 201 conforms to the rules of the specified syntax. The bit string decoding unit 201 decodes the separated first bit string and Sequence, picture, slice, coding block unit information and coding block unit Specifically, the bitstream decoding unit 201 obtains the coded information by inter-coding block units. Prediction mode PredMode determines whether the prediction is intra prediction (MODE_INTER) or intra prediction (MODE_INTRA). When the prediction mode is inter prediction (MODE_INTER), the bitstream decoding unit 201 Flag to determine whether it is in merge mode, merge index if in merge mode, sub Block merge flag, inter prediction mode if predicted motion vector mode, prediction Motion vector index, differential motion vector, sub-block predicted motion vector flag, etc. The coding information (inter prediction information) is decoded according to a specified syntax and coded. information (inter prediction information) via the inter prediction unit 203 and the block division unit 202 The prediction mode is intra prediction (MODE_INTRA). In this case, coding information (intra prediction information) such as intra prediction mode is specified in the specified syntax. Therefore, the encoded information (intra prediction information) is decoded and transmitted to the inter prediction unit 203 or the intra prediction unit 204. The data is supplied to the coding information storage memory 205 via the prediction unit 204 and the block division unit 202. The bit stream decoding unit 201 decodes the separated second bit stream and performs orthogonal transform and quantization. The orthogonally transformed and quantized residual is supplied to the inverse quantization and inverse orthogonal transformation unit 206. do.

[0042] The inter prediction unit 203 performs the inter prediction when the prediction mode PredMode of the coding block to be processed is set to inter prediction. When the prediction (MODE_INTER) is in the predicted motion vector mode, the coding information storage memory 205 A plurality of predicted motion vectors are generated using the coding information of the already decoded image signal stored in the The motion vector predictor candidates are derived, and the derived motion vector predictor candidates are used as the motion vector predictor candidates described later. The inter prediction unit 203 registers the motion vector predicted candidate in the motion vector predicted candidate list. From among the plurality of motion vector predictor candidates recorded, a bitstream decoder 201 decodes and supplies a motion vector predictor corresponding to the motion vector predictor index, and decoding the bitstream; The motion vector is calculated from the differential motion vector decoded by the unit 201 and the selected predicted motion vector. The calculated motion vector is stored in the coded information storage memory 205 together with other coded information. The coding information of the coding block to be supplied and stored here is the prediction mode PredMode , flags predFlagL0[xP][yP], predFlag indicating whether to use L0 prediction and L1 prediction L1[xP][yP], L0, L1 reference indices refIdxL0[xP][yP], refIdxL1[xP][yP], L0 , L1 motion vectors mvL0[xP][yP], mvL1[xP][yP], etc., where xP and yP are the The Prediction Mode PredMo is an index indicating the position of the top left pixel of the coding block in the Prediction Mode PredMo. When de is inter prediction (MODE_INTER) and the inter prediction mode is L0 prediction (Pred_L0) , the flag predFlagL0 indicating whether to use L0 prediction is 1, and the flag predFlagL0 indicating whether to use L1 prediction is 2. The flag predFlagL1 indicating this is 0. When the inter prediction mode is L1 prediction (Pred_L1), The flag predFlagL0 indicating whether to use L0 prediction is 0, and the flag predFlagL1 indicating whether to use L1 prediction is 1. The flag predFlagL1 is set to 1. When the inter prediction mode is bi-prediction (Pred_BI), Flag indicating whether prediction is used or not predFlagL0, Flag indicating whether L1 prediction is used or not predFlagL1 and predFlagL2 are both set to 1. Furthermore, the prediction mode PredMode of the coding block to be processed is When merging in inter prediction (MODE_INTER) mode, merging candidates are derived. Using the coding information of the already decoded coding block stored in the storage memory 205 , multiple merge candidates are derived and registered in a merge candidate list described later. The merge candidates are decoded by the bitstream decoder 201 and supplied from among the multiple merge candidates registered in the Select a merge candidate corresponding to the bin index, and perform L0 prediction of the selected merge candidate. Flags predFlagL0[xP][yP], predFlagL1[xP][yP], and L0 , the reference index of L1 refIdxL0[xP][yP], refIdxL1[xP][yP], the motion vector of L0, L1 Inter prediction information such as mvL0[xP][yP] and mvL1[xP][yP] is stored in the encoding information storage memory 205. Here, xP and yP are the indexes that indicate the position of the top left pixel of the coding block in the picture. The detailed configuration and operation of the inter prediction unit 203 will be described later.

[0043] The intra prediction unit 204 determines whether the prediction mode PredMode of the coding block to be processed is intra. In the case of prediction (MODE_INTRA), intra prediction is performed. The encoding information includes an intra-prediction mode. The decoding unit 201 decodes the image data according to the intra prediction mode included in the coded information. A predicted image signal is generated by intra-prediction from the decoded image signal stored in the memory 208. The generated predicted image signal is supplied to the decoded image signal superimposing unit 207. 204 corresponds to the intra prediction unit 103 of the image encoding device 100. The same processing as that of the intra prediction unit 103 is performed.

[0044] The inverse quantization and inverse orthogonal transformation unit 206 performs orthogonal transformation and quantization on the decoded bit stream by the bit stream decoding unit 201. The inverse orthogonal transform and inverse quantization are performed on the residual, and the inverse orthogonal transformed and inverse quantized residual is get.

[0045] The decoded image signal superimposing unit 207 superimposes the predicted image signal obtained by the inter prediction unit 203. or a predicted image signal obtained by intra-prediction in the intra prediction unit 204, and The decoded image is obtained by superimposing the residuals that have been inverse orthogonally transformed and inversely quantized by the orthogonal transform unit 206. The decoded image signal is stored in the decoded image memory 208. When storing the decoded image in the memory 208, the decoded image signal superimposing unit 207 performs coding on the decoded image. After filtering to reduce block distortion etc., the decoded image is stored in the decoded image memory 208. It may be stored.

[0046] Next, the operation of the block division unit 101 in the image coding device 100 will be described. Figure 3 shows the division of an image into treeblocks and the further division of each treeblock. First, the input image is divided into tree blocks of a predetermined size. Each tree block is sorted in a predetermined order, i.e., a raster order (step S1001). The tree block is scanned in scan order (step S1002) and the inside of the tree block to be processed is divided. (Step S1003).

[0047] FIG. 7 is a flowchart showing the detailed operation of the division process in step S1003. It is determined whether or not the block to be processed is to be divided into four (step S1101).

[0048] If it is determined that the processing block should be divided into four, the processing block is divided into four (step Step S1102). For each block into which the processing target block is divided, That is, the scanning is performed in the order of upper left, upper right, lower left, and lower right (step S1103). 6A is an example of a scan order, and 601 in FIG. 6A is an example in which the block to be processed is divided into four. The numbers 0 to 3 in 601 of A indicate the order of processing. For each divided block, the division process of FIG. 7 is recursively executed (step S1104 ).

[0049] If it is determined that the block to be processed is not to be divided into four, it is divided into two or three (step S1 105).

[0050] FIG. 8 is a flowchart showing the detailed operation of the 2-3 division process in step S1105. First, we decide whether to divide the block to be processed into 2 or 3 parts. It is determined whether or not to perform either of these (step S1201).

[0051] If it is not determined that the processing block should be divided into 2-3 blocks, i.e., if it is determined that no division should be made, In other words, the division is completed by the recursive division process (step S1211). No further recursive division processing is performed on the blocks that have been divided.

[0052] If it is determined that the processing block should be divided into 2-3 blocks, the processing block is further divided into 2 blocks. It is determined whether or not to divide (step S1202).

[0053] If it is determined that the processing block is to be divided into two, the processing block is divided into upper and lower (vertical ) (step S1203), and based on the result, the block to be processed is The block is divided into two parts vertically (step S1204), or the block to be processed is divided into two parts vertically (step S1205). Divide the image into two on the right (horizontal direction) (step S1205). The target block is divided into two parts, top and bottom (vertical direction), as shown in 602 of FIG. 6B, and the steps are As a result of the step S1205, the block to be processed is divided into left and right (horizontal) blocks as shown in 604 of FIG. 6D. (Directional) It is divided into two parts.

[0054] In step S1202, if it is not determined that the block to be processed is to be divided into two In other words, if it is determined that the block should be divided into three, the block to be processed is divided into top, middle and bottom (vertical direction). Based on the result of the determination, the block to be processed is moved up or down (step S1206). Divide the target block into three parts (vertically) (step S1207), or divide the target block into three parts (left, middle, right, and As a result of step S1207, the processing object is divided into three parts (horizontal direction) (step S1208). The block is divided into three parts (vertical direction) - top, middle and bottom, as shown in 603 of FIG. 6C. As a result of step S1208, the block to be processed is divided into left, center, and right (horizontal) blocks as shown in 605 of FIG. 6E. (Directional) It is divided into three parts.

[0055] Step S1204, Step S1205, Step S1207, Step S1208 After performing either of the above, for each block into which the processing target block is divided, The numbers 0 to 605 in FIGS. 2 shows the order of processing. For each divided block, 2-3 division processing in Fig. 8 is performed. The process is recursively executed (step S1210).

[0056] The recursive block division explained here is performed by dividing the number of blocks into multiple blocks. The necessity of division may be restricted depending on the size of the image. By making a prearrangement between the decoder and the decoder, it is possible to realize a configuration that does not transmit information. Alternatively, the encoding device may determine information that limits the necessity of division and record it in the bit string. , may be realized by transmitting the information to the decoding device.

[0057] When a block is divided, the block before the division is called the parent block, and each block after the division is called the parent block. The blocks are called child blocks.

[0058] Next, the operation of the block division unit 202 in the image decoding device 200 will be described. The block division unit 202 performs the same processing procedure as the block division unit 101 of the image encoding device 100. However, the block division of the image coding device 100 is The unit 101 applies optimization techniques such as optimal shape estimation by image recognition and strain rate optimization, The optimal block division shape is determined, whereas the block division in the image decoding device 200 is determined. The dividing unit 202 divides the blocks by decoding the block division information recorded in the bit string. The difference is that the division shape is determined.

[0059] The syntax (syntax rules for bit strings) related to block division in the first embodiment is shown in FIG. coding_quadtree() represents the syntax for dividing a block into four parts. _type_tree() represents the syntax for splitting a block into two or three parts. it is a flag that indicates whether to divide the block into four parts. If you want to divide the block into four parts, use qt If you do not want to split into four, set qt_split=0. If you want to split into four (qt_split=1), For each block divided into four, recursively divide it into four (coding_quadtree(0), codin g_quadtree(1), coding_quadtree(2), coding_quadtree(3), arguments 0 to 3 correspond to the 6 in Figure 6A. (This corresponds to the number 01.) If you do not split into four (qt_split=0), follow the multi_type_tree() , determines the subsequent split. mtt_split is a flag indicating whether to split further. If further splitting is required (mtt_split=1), indicate whether to split vertically or horizontally. The flag mtt_split_vertical determines whether to split the image into two or three parts. mtt_split_vertical=1 indicates splitting in the vertical direction. mtt_split_vertical=0 indicates splitting horizontally, and mtt_split_binary=1 indicates splitting horizontally. mtt_split_binary=0 indicates that it will be split into two, and mtt_split_binary=0 indicates that it will be split into three. tt_split_binary=1), and recursively split each block into two (multi_t type_tree(0), multi_type_tree(1), arguments 0 to 1 correspond to 602 or 604 in Figure 6B to D. (These numbers correspond to the numbers.) When splitting into three (mtt_split_binary=0), the following is true for each of the three split blocks: and recursively splits the tree (multi_type_tree(0), multi_type_tree(1), multi_type_ tree(2), 0 to 2 correspond to the numbers 603 in Figure 6B or 605 in Figure 6E.) mtt_spli Hierarchical block division by recursively calling multi_type_tree until t=0. Do the following.

[0060] <Inter prediction> The inter prediction method according to the embodiment is performed by the inter prediction unit 102 of the image encoding device shown in FIG. This is performed in the inter prediction unit 203 of the image decoding device in FIG.

[0061] The inter prediction method according to the embodiment will be described with reference to the drawings. The method is performed in both the encoding and decoding processes on a coding block basis.

[0062] <Explanation of the Inter Prediction Unit 102 on the Encoding Side> FIG. 16 is a diagram showing a detailed configuration of the inter prediction unit 102 of the image encoding device of FIG. The normal motion vector predictor mode derivation unit 301 derives a plurality of normal motion vector predictor candidates. A predicted motion vector is selected by the motion vector prediction function, and the selected predicted motion vector is compared with the detected motion vector. The differential motion vector is calculated based on the detected inter prediction mode, reference index, and motion vector. The calculated differential motion vector is the inter prediction in normal prediction motion vector mode. This inter prediction information is supplied to the inter prediction mode determination unit 305. The detailed configuration and processing of the predictive motion vector mode derivation unit 301 will be described later.

[0063] The normal merge mode derivation unit 302 derives multiple normal merge candidates and This inter prediction information is used for the normal merge mode. The result is supplied to the prediction mode determination unit 305. The reasoning behind this will be explained later.

[0064] The sub-block prediction motion vector mode derivation unit 303 calculates a plurality of sub-block prediction motion vectors. A motion vector candidate is derived to select a sub-block predictor motion vector, and the selected sub-block predictor motion vector is A differential motion vector is calculated between the estimated motion vector and the detected motion vector. Inter prediction mode, reference index, motion vector, calculated differential motion vector This becomes inter prediction information for the sub-block prediction motion vector mode. is supplied to the inter prediction mode determination unit 305.

[0065] The sub-block merge mode derivation unit 304 derives multiple sub-block merge candidates. Select sub-block merging candidates and obtain inter-prediction information for sub-block merging modes This inter prediction information is supplied to the inter prediction mode determination unit 305.

[0066] The inter prediction mode determination unit 305 determines whether the normal prediction motion vector mode is derived from the normal prediction motion vector mode derivation unit 301 or the normal prediction motion vector mode derivation unit 302. A merge mode derivation unit 302, a sub-block predicted motion vector mode derivation unit 303, a sub-block Based on the inter prediction information supplied from the block merge mode derivation unit 304, The inter prediction mode determination unit 305 determines the prediction information. The measurement information is supplied to the motion compensation prediction unit 306 .

[0067] The motion compensation prediction unit 306 stores the decoded image in the decoded image memory 1 based on the determined inter prediction information. The motion compensation prediction unit 306 performs inter prediction on the reference image signal stored in the motion compensation prediction unit 306. The detailed configuration and processing will be described later.

[0068] <Description of the Inter Prediction Unit 203 on the Decoding Side> FIG. 22 is a diagram showing a detailed configuration of the inter prediction unit 203 of the image decoding device of FIG.

[0069] The normal motion vector predictor mode derivation unit 401 derives a plurality of normal motion vector predictor candidates. a predicted motion vector is selected, and the selected predicted motion vector and the decoded differential motion vector are The decoded inter prediction mode, reference index, and In this case, the motion vector is the inter prediction information in the normal prediction motion vector mode. The center prediction information is supplied to the motion compensation prediction unit 406 via the switch 408. The detailed configuration and processing of the motion vector measurement mode derivation unit 401 will be described later.

[0070] The normal merge mode derivation unit 402 derives multiple normal merge candidates and selects the normal merge candidates. Select this to get the inter prediction information for normal merge mode. The merge mode derivation unit 402 normally supplies the merge mode to the motion compensation prediction unit 406 via 408. The detailed configuration and processing will be described later.

[0071] The sub-block prediction motion vector mode derivation unit 403 calculates a plurality of sub-block prediction motion vectors. A motion vector candidate is derived to select a sub-block predictor motion vector, and the selected sub-block predictor motion vector is The sum of the estimated motion vector and the decoded differential motion vector is calculated and used as the motion vector. The decoded inter prediction mode, reference index, and motion vector are used for sub-block prediction. This inter prediction information is used as the inter prediction information for the motion vector mode. is supplied to the motion compensation prediction unit 406 via

[0072] The sub-block merge mode derivation unit 404 derives multiple sub-block merge candidates. Select sub-block merging candidates and obtain inter-prediction information for sub-block merging modes This inter prediction information is supplied to the motion compensation prediction unit 406 via the switch 408. .

[0073] The motion compensation prediction unit 406 stores the decoded image data in the decoded image memory 2 based on the determined inter-prediction information. 08. The motion compensation prediction unit 406 performs inter-prediction on the reference image signal stored in the motion compensation prediction unit 406. The detailed configuration and processing of the motion compensation prediction unit 306 on the encoding side are the same as those of the motion compensation prediction unit 306 on the encoding side.

[0074] <Normal predicted motion vector mode derivation part (normal AMVP)> The normal predicted motion vector mode derivation unit 301 in FIG. 17 derives spatial predicted motion vector candidates. 321, a temporal motion vector predictor candidate derivation unit 322, and a historical motion vector predictor candidate derivation unit 3 23, a predicted motion vector candidate supplementation unit 325, a normal motion vector detection unit 326, a predicted motion vector The motion vector selection unit 327 includes a motion vector candidate selection unit 328 .

[0075] The normal predicted motion vector mode derivation unit 401 in FIG. 23 derives spatial predicted motion vector candidates. unit 421, a temporal motion vector predictor candidate derivation unit 422, and a historical motion vector predictor candidate derivation unit 4 23, a motion vector predictor candidate supplementation unit 425, a motion vector predictor candidate selection unit 426, a motion vector The vector adder 427 is included.

[0076] The normal predicted motion vector mode derivation unit 301 on the encoding side and the normal predicted motion vector mode derivation unit 302 on the decoding side The processing procedure of the torque mode derivation unit 401 is shown in the flowcharts of FIGS. 19 and 25. 19 is a diagram showing the normal motion vector mode derivation unit 301 on the encoding side. 25 is a flowchart showing a procedure for deriving a predicted motion vector mode. 1 shows a procedure for deriving a normal predicted motion vector mode by the normal motion vector mode derivation unit 401. This is a flowchart.

[0077] <Normal prediction motion vector mode derivation part (normal AMVP): explanation on the encoding side> The normal predicted motion vector mode derivation process on the encoding side will be described with reference to FIG. In the explanation of the processing procedure of FIG. 19, the word "normal" shown in FIG. 19 may be omitted.

[0078] First, the normal motion vector detection unit 326 detects the inter prediction mode and the reference index. Then, a normal motion vector is detected (step S100 in FIG. 19).

[0079] Next, the spatial motion vector predictor candidate derivation unit 321 and the temporal motion vector predictor candidate derivation unit 3 22, a history predicted motion vector candidate derivation unit 323, a predicted motion vector candidate supplement unit 325, a predicted motion vector candidate A predicted motion vector candidate selection unit 327 and a motion vector subtraction unit 328 select a normal predicted motion vector The differential motion vectors of the motion vectors used in inter prediction of the mode are set for L0 and L1 respectively. Specifically, the predicted value of the block to be processed is calculated (steps S101 to S106 in FIG. 19). Prediction mode PredMode is inter prediction (MODE_INTER), and inter prediction mode is L0 prediction (Pr ed_L0), calculate the motion vector predictor candidate list mvpListL0 of L0, and The motion vector mvpL0 of L0 is selected, and the differential motion vector mvdL0 of the motion vector mvL0 of L0 is calculated. If the inter prediction mode of the target block is L1 prediction (Pred_L1), the predicted motion vector of L1 is A motion vector candidate list mvpListL1 is calculated, a motion vector predictor mvpL1 is selected, and the motion vector of L1 is calculated. Calculate the differential motion vector mvdL1 of the current block mvL1. When bi-prediction (Pred_BI) is used, both L0 prediction and L1 prediction are performed, and the predicted motion vector of L0 is A motion vector candidate list mvpListL0 is calculated, a motion vector predictor mvpL0 for L0 is selected, and the motion vector A differential motion vector mvdL0 of the vector mvL0 is calculated, and a predicted motion vector candidate of L1 is calculated. A complementary list mvpListL1 is calculated, and a predicted motion vector mvpL1 of L1 is calculated. Then, the differential motion vector mvdL1 of the corresponding vector mvL1 is calculated.

[0080] The differential motion vector calculation process is performed for each of L0 and L1. Therefore, in the following explanation, L0 and L1 are represented as a common LX. In the process of calculating the differential motion vector of L0, X of LX is 0, and In the process of calculating the motion vector of LX, X of LX is 1. If you refer to the information in the other list instead of the LX during the process of adding the other list, is expressed as LY.

[0081] When the motion vector mvLX of LX is used (step S102 in FIG. 19: YES), Calculate the candidate predicted motion vectors of LX and construct a candidate predicted motion vector list mvpListLX of LX. (Step S103 in FIG. 19). A spatial motion vector predictor candidate derivation unit 321, a temporal motion vector predictor candidate derivation unit 322, a history A motion vector predictor candidate deriving unit 323 and a motion vector predictor candidate supplementing unit 325 generate a plurality of motion vector predictors. The motion vector predictor candidate list mvpListLX is constructed by deriving motion vector candidates. The detailed processing procedure of step S103 will be described later with reference to the flowchart of FIG. do.

[0082] Next, the motion vector predictor candidate selection unit 327 selects a motion vector predictor candidate list for LX. Select a predicted motion vector mvpLX for LX from the mvpListLX (step S104 in FIG. 19). Here, in the motion vector predictor candidate list mvpListLX, a certain element (counting from 0) The motion vector mvLX and the motion vector predictor candidate are expressed as mvpListLX[i]. The difference between each candidate motion vector predictor mvpListLX[i] stored in the list mvpListLX is The differential motion vectors are calculated for each of the coded frames. The coding amount is calculated for each element (predictor motion vector candidate) of the predictor motion vector candidate list mvpListLX. Then, among the elements registered in the motion vector predictor candidate list mvpListLX, The candidate motion vector predictor mvpListLX[i] that minimizes the amount of coding for each candidate motion vector predictor is The motion vector predictor mvpLX is selected as the motion vector predictor mvpLX, and its index i is obtained. There are multiple candidates for the predicted motion vector with the minimum amount of generated code in the candidate list mvpListLX. In this case, the index i in the motion vector predictor candidate list mvpListLX is a small number. The candidate motion vector predictor mvpListLX[i] represented by the following is used as the optimal motion vector predictor mvpLX. Select it and get its index i.

[0083] Next, the motion vector subtraction unit 328 subtracts the motion vector mvLX of LX selected from the motion vector mvLX of LX. Subtract the predicted motion vector mvpLX, mvdLX = mvLX - mvpLX The differential motion vector mvdLX of LX is calculated as follows (step S105 in FIG. 19).

[0084] <Normal predicted motion vector mode derivation unit (normal AMVP): Decoding side explanation> Next, the normal predicted motion vector mode processing procedure on the decoding side will be described with reference to FIG. On the other hand, a spatial prediction motion vector candidate derivation unit 421 and a temporal prediction motion vector candidate derivation unit 422 are provided. 22, the history predicted motion vector candidate derivation unit 423, the predicted motion vector candidate supplementation unit 425, The motion vectors used in inter prediction in normal prediction motion vector mode are set for each of L0 and L1. Specifically, the process of the block to be processed is performed as follows (steps S201 to S206 in FIG. 25). Prediction mode PredMode is inter prediction (MODE_INTER) and the target block is inter prediction When the mode is L0 prediction (Pred_L0), the predicted motion vector candidate list mvpListL0 of L0 is calculated. Then, the predicted motion vector mvpL0 is selected, and the motion vector mvL0 of L0 is calculated. If the inter prediction mode of the target block is L1 prediction (Pred_L1), the predicted motion vector of L1 is used. A motion vector candidate list mvpListL1 is calculated, a motion vector predictor mvpL1 is selected, and the motion vector of L1 is If the inter prediction mode of the block to be processed is bi-predictive (Pred_BI), Both L0 prediction and L1 prediction are performed, and the L0 predicted motion vector candidate list mvpListL0 is calculated. Then, the predicted motion vector mvpL0 of L0 is selected, and the motion vector mvL0 of L0 is calculated. In both cases, a motion vector predictor candidate list mvpListL1 of L1 is calculated, and the motion vector predictor of L1 is Then, the motion vector mvpL1 for L1 is calculated, and the motion vector mvL1 for L2 is calculated.

[0085] As with the encoding side, the decoding side also performs motion vector calculation processing for each of L0 and L1. However, the process is common to both L0 and L1. L1 is represented as a common LX. LX is used for inter prediction of the coding block being processed. This represents the inter prediction mode. In the process of calculating the L0 motion vector, X is 0, and in the L1 In the process of calculating the motion vector of L, X is 1. During processing, the LX being calculated is not based on the same reference list, but on information from another reference list. When referencing one reference list, the other reference list is represented as LY.

[0086] When the motion vector mvLX of LX is used (step S202 in FIG. 25: YES), LX Calculate the candidate predicted motion vectors of LX and construct a candidate predicted motion vector list mvpListLX of LX. (Step S203 in FIG. 25). A spatial motion vector predictor candidate derivation unit 421, a temporal motion vector predictor candidate derivation unit 422, a history A motion vector predictor candidate deriving unit 423 and a motion vector predictor candidate supplementing unit 425 generate a plurality of motion vector predictors. Then, motion vector predictor candidates are calculated and a motion vector predictor candidate list mvpListLX is constructed. The detailed processing procedure of step S203 will be described later with reference to the flowchart of FIG. do.

[0087] Next, the motion vector predictor candidate selection unit 426 selects the motion vector predictor candidate list mvpListLX The bitstream decoder 201 decodes and supplies the predicted motion vector index mv The candidate predicted motion vector mvpListLX[mvpIdxLX] corresponding to pIdxLX is used as the selected predicted motion vector. The vector mvpLX is extracted (step S204 in FIG. 25).

[0088] Then, the motion vector adder 427 decodes the bit stream and supplies it to the bit stream decoder 201. Add the differential motion vector mvdLX of LX and the predicted motion vector mvpLX of LX, mvLX = mvpLX + mvdLX The motion vector mvLX of LX is calculated as follows (step S205 in FIG. 25).

[0089] <Normal predicted motion vector mode derivation part (normal AMVP): Motion vector prediction method> FIG. 20 shows a normal prediction motion vector mode derivation process of the image coding apparatus according to the embodiment of the present invention. Functions common to the motion vector mode derivation unit 301 and the normal predicted motion vector mode derivation unit 401 of the image decoding device 10 is a flowchart showing the processing procedure of a normal predicted motion vector mode derivation process having the following formula.

[0090] The normal predicted motion vector mode derivation unit 301 and the normal predicted motion vector mode derivation unit 40 1 includes a motion vector predictor candidate list mvpListLX. The mvpListLX has a list structure and is a prediction vector that indicates the location within the motion vector prediction candidate list. The motion vector index and the motion vector predictor candidate corresponding to the index are used as elements. The predicted motion vector index number starts from 0. The motion vector predictor candidate list mvpListLX is started and stored in the storage area of ​​the motion vector predictor candidate list mvpListLX. In this embodiment, the motion vector predictor candidate list mvpListLX contains at least At least two candidate motion vector predictors (inter prediction information) can be registered. Furthermore, the motion vector predictor registered in the motion vector predictor candidate list mvpListLX is The variable numCurrMvpCand, which indicates the number of rule candidates, is set to 0.

[0091] The spatial motion vector predictor candidate derivation units 321 and 421 derive the spatial motion vector predictor candidate from the adjacent block on the left side. In this process, candidates for predicted motion vectors are derived from the adjacent block on the left ( A0 or A1) inter prediction information, i.e., whether or not a motion vector predictor candidate is available. A flag indicating whether the motion vector is a predicted motion vector mv LXA is derived, and the derived mvLXA is added to the motion vector predictor candidate list mvpListLX (FIG. 20 In the case of L0 prediction, X is set to 0, and in the case of L1 prediction, X is set to 1 (hereinafter, Next, the spatial prediction motion vector candidate derivation units 321 and 421 select the adjacent spatial prediction motion vector candidate on the upper side. In this process, the candidate predicted motion vectors are derived from the adjacent block on the upper side. Inter prediction information of the block (B0, B1, or B2 in FIG. 11), i.e., predicted motion vector A flag indicating whether a motion candidate is available or not, and a reference to the motion vector, reference index, etc. The predicted motion vector mvLXA and mvLXB are derived based on the calculated motion vector mvLXA and mvLXB. If so, mvLXB is added to the motion vector predictor candidate list mvpListLX (step S3 in FIG. 20). 02). The processing in steps S301 and S302 in FIG. 20 is performed by determining the position and number of adjacent blocks to be referenced. The difference is whether the candidate predicted motion vector for the coding block can be used. The flag availableFlagLXN indicates whether the motion vector is available or not, and the motion vector mvLXN and the reference index refIdxN( N indicates A or B, and so on.

[0092] Next, the temporal motion vector predictor candidate derivation units 322 and 422 calculate the temporal motion vector predictor candidate for the current processing target picture. A candidate motion vector predictor is derived from a block in a picture that is different in time from the image. In this process, motion vector predictor candidates of coding blocks of pictures at different times are used. The flag availableFlagLXCol indicates whether the motion vector mvLXCol and the reference index are available. The reference list refIdxCol and the reference list listCol are derived, and mvLXCol is used to create the predicted motion vector candidate list m Add it to vpListLX (step S303 in FIG. 20).

[0093] Note that temporal prediction is performed in units of sequences (SPS), pictures (PPS), or slices. It is assumed that the processing of the motion vector candidate derivation units 322 and 422 can be omitted.

[0094] Next, the history prediction motion vector candidate derivation units 323 and 423 derive the history prediction motion vector candidate. The historical predicted motion vector candidates registered in the supplementary list HmvpCandList are used as predicted motion vector candidates. The mvpListLX is added to the auxiliary list mvpListLX (step S304 in FIG. 20). The registration process will be described in detail later with reference to the flowchart of FIG.

[0095] Next, the motion vector predictor candidate supplementation units 325 and 425 generate the motion vector predictor candidate list mv Add motion vector predictor candidates with a predetermined value, such as (0,0), until pListLX is satisfied ( S305 in Figure 20).

[0096] <Normal merge mode derivation part (normal merge)> The normal merge mode derivation unit 302 in FIG. 18 includes a spatial merge candidate derivation unit 341, a temporal merge candidate derivation unit 342, and a Candidate derivation unit 342, average merge candidate derivation unit 344, history merge candidate derivation unit 345, merge It includes a candidate supplementation unit 346 and a merge candidate selection unit 347 .

[0097] The normal merge mode derivation unit 402 in FIG. 24 includes a spatial merge candidate derivation unit 441, a temporal merge candidate derivation unit 442, and a Candidate derivation unit 442, average merge candidate derivation unit 444, history merge candidate derivation unit 445, merge It includes a candidate supplementation unit 446 and a merge candidate selection unit 447 .

[0098] FIG. 21 shows the normal merge mode derivation unit 302 and A normal merge mode having a function common to the normal merge mode derivation unit 402 of the image decoding device and the normal merge mode derivation unit 402 of the image decoding device. 10 is a flowchart illustrating the procedure of a code derivation process.

[0099] The following describes the various steps in order. The slice type slice_type is explained as a B slice, but the case of a P slice is explained as follows. However, if the slice type slice_type is P slice, the inter prediction model There is only L0 prediction (Pred_L0) as a code, and L1 prediction (Pred_L1) and bi-prediction (Pred_BI) are also supported. Since there is no L1, processing related to L1 can be omitted.

[0100] The normal merge mode derivation unit 302 and the normal merge mode derivation unit 402 determine the merge candidate list. The merge candidate list mergeCandList has a list structure, A merge index that indicates the location within the merge candidate list and the merge candidate corresponding to the index. There is a memory area for storing merge candidates as elements. The merge index number is 0. The merge candidates are stored in the merge candidate list (mergeCandList) starting from In the following process, the merge index i registered in the merge candidate list mergeCandList is used. The merge candidates are represented by mergeCandList[i]. The candidate list mergeCandList must contain at least six merge candidates (inter-prediction information). Furthermore, the merge candidates registered in the mergeCandList The variable numCurrMergeCand, which indicates the number of merge candidates, is set to 0.

[0101] The spatial merge candidate derivation unit 341 and the spatial merge candidate derivation unit 441 are The encoded information stored in the encoded information storage memory 111 or the encoded information storage memory 205 of the image decoding device From the coding information stored in the block, each block adjacent to the block to be processed (B in FIG. 11) 1, A1, B0, A0, B2) to B1, A1, B0, A0, B2 The spatial merge candidates are derived in order and registered in the merge candidate list mergeCandList. (Step S401 in FIG. 21). Here, B1, A1, B0, A0, B2 or time marker We define N, which indicates one of the candidate blocks Col. A flag availableFlagN indicates whether the candidate can be used as a spatial merge candidate. Reference index refIdxL0N and L1 reference index refIdxL1N, L0 prediction is performed L0 prediction flag predFlagL0N indicates whether L1 prediction is performed or not, and L1 prediction flag predFlagL0N indicates whether L1 prediction is performed or not. The motion vector mvL0N of L0 and the motion vector mvL1N of L1 are derived. However, in this embodiment, the blocks included in the coding block to be processed are Since merge candidates are derived without referring to inter-prediction information, The spatial merge candidates using the inter prediction information of the blocks included in the matrix B are not derived.

[0102] Next, the temporal merge candidate derivation unit 342 and the temporal merge candidate derivation unit 442 determine different time derive temporal merge candidates from the pictures between, and the derived temporal merge candidates are called merge candidates The time merge candidate is registered in the list mergeCandList (step S402 in FIG. 21). flag availableFlagCol indicating whether L0 prediction of temporal merge candidates is performed or not. L0 prediction flag predFlagL0Col indicating whether L1 prediction is performed and L1 prediction flag predFlagL0Col indicating whether L1 prediction is performed. The flag predFlagL1Col, the motion vector mvL0Col of L0, and the motion vector mvL1Col of L1 are Derive.

[0103] Note that time markers can be set in units of sequences (SPS), pictures (PPS), or slices. The processing of the merge candidate derivation unit 342 and the temporal merge candidate derivation unit 442 can be omitted. Let's say.

[0104] Next, the history merge candidate derivation unit 345 and the history merge candidate derivation unit 445 perform history prediction. Mark the historical predicted motion vector candidates registered in the motion vector candidate list HmvpCandList. The candidate is registered in the mergeCandList (step S403 in FIG. 21). The number of merge candidates registered in the merge candidate list mergeCandList is numCurrMergeC If and is smaller than the maximum number of merge candidates MaxNumMergeCand, the merge candidate list mergeCandL The number of merge candidates registered in ist is numCurrMergeCand, and the maximum number of merge candidates is MaxNumMergeCand. The history merge candidates are derived with nd as the upper limit and registered in the merge candidate list mergeCandList. can be.

[0105] Next, the average merge candidate derivation unit 344 and the average merge candidate derivation unit 444 calculate the merge candidates. The average merge candidate is derived from the complement list mergeCandList, and the derived average merge candidate is merged. The merge candidate list mergeCandList is then added to the mergeCandList (step S404 in FIG. 21). The number of merge candidates registered in the merge candidate list mergeCandList is numCurrMergeC If and is smaller than the maximum number of merge candidates MaxNumMergeCand, the merge candidate list mergeCandL The number of merge candidates registered in ist is numCurrMergeCand, and the maximum number of merge candidates is MaxNumMergeCand. The average merge candidate is calculated with nd as the upper limit and registered in the merge candidate list mergeCandList. can be. Here, the average merge candidate is the first merge candidate registered in the merge candidate list mergeCandList. The motion vectors of the merge candidate and the second merge candidate are averaged for each of the L0 prediction and the L1 prediction. The resulting motion vector is a new merge candidate.

[0106] Next, the merge candidate supplementation unit 346 and the merge candidate supplementation unit 446 generate a merge candidate list. The number of merge candidates registered in mergeCandList, numCurrMergeCand, is less than the maximum number of merge candidates, M If it is smaller than axNumMergeCand, the merge candidate list mergeCandList contains The number of merge candidates numCurrMergeCand is the maximum number of merge candidates MaxNumMergeCand. The merge candidates are derived and registered in the merge candidate list mergeCandList (step S4 in FIG. 21). 05). In P slices, the maximum number of merge candidates is MaxNumMergeCand. A merge candidate having a prediction mode of L0 prediction (Pred_L0) with a value of (0,0) is added. In B slices, the prediction mode with a motion vector of (0,0) is bi-predictive (Pred_BI). The reference index when adding a merge candidate is the same as the index of the merge candidate already added. This is different from the reference index used.

[0107] Next, the merge candidate selection unit 347 and the merge candidate selection unit 447 select a merge candidate list Select a merge candidate from the merge candidates registered in mergeCandList. The merge candidate selection unit 347 selects merge candidates by calculating the code amount and distortion amount. , a merge index indicating the selected merge candidate, inter prediction information of the merge candidate, The inter-prediction mode determination unit 305 supplies the inter-prediction mode data to the motion compensation prediction unit 306. The merge candidate selection unit 447 on the side selects merge candidates based on the decoded merge index. The motion compensation prediction unit 406 selects a merge candidate and supplies the selected merge candidate to the motion compensation prediction unit 406 .

[0108] <Sub-block prediction motion vector mode derivation> The sub-block predicted motion vector mode derivation will be explained. FIG. 38 shows the sub-block predicted motion vector mode derivation unit 30 in the encoding device of the present application. 3 is a block diagram of the

[0109] First, the affine inheritance predicted motion vector candidate derivation unit 361 derives the affine inheritance predicted motion vector candidate. The details of deriving affine inheritance predicted motion vector candidates will be explained later. Describe.

[0110] Next, the affine construction prediction motion vector candidate derivation unit 362 calculates the affine construction prediction Derive motion vector candidates. For details on deriving affine construction prediction motion vector candidates, see This will be discussed later.

[0111] Next, the affine-coherent prediction motion vector candidate derivation unit 363 performs affine-coherent prediction. Derive motion vector candidates. For details on deriving affine identical predictor motion vector candidates, see This will be discussed later.

[0112] The sub-block motion vector detector 366 is adapted to the sub-block predicted motion vector mode. A sub-block motion vector corresponding to the sub-block is detected, and the detected vector is used as a sub-block predicted motion vector. The result is supplied to a torque candidate selection unit 367 and a difference calculation unit 368.

[0113] The sub-block predicted motion vector candidate selection unit 367 selects the affine inheritance predicted motion vector candidate. A complementary derivation unit 361, an affine construction predicted motion vector candidate derivation unit 362, an affine identical predicted motion vector Among the sub-block predicted motion vector candidates derived by the motion vector candidate derivation unit 363, Based on the motion vector supplied from the sub-block motion vector detection unit 366, A sub-block motion vector predictor candidate is selected, and the selected sub-block motion vector predictor The information about the candidates is supplied to the inter prediction mode determination unit 305 and the difference calculation unit 368 .

[0114] The difference calculation unit 368 calculates the motion vectors supplied from the sub-block motion vector detection unit 366. The sub-block predicted motion vector candidate selection unit 367 selects the sub-block predicted motion vector candidate from the motion vector. The difference prediction motion vector obtained by subtracting the inter prediction motion vector is Supply to 305.

[0115] FIG. 39 shows the sub-block predicted motion vector mode derivation unit 403 in the decoding device of the present application. FIG.

[0116] First, the affine inheritance predicted motion vector candidate derivation unit 461 derives the affine inheritance predicted motion vector candidate. The processing of the affine inheritance predicted motion vector candidate derivation unit 461 is The processing is the same as that of the affine inheritance predicted motion vector candidate derivation unit 361 in the encoding device of the present application. be.

[0117] Next, the affine construction prediction motion vector candidate derivation unit 462 calculates the affine construction prediction The affine construction prediction motion vector candidate derivation unit 462 performs the following process: The same processing as that of the affine construction predicted motion vector candidate derivation unit 362 in the encoding device of the present application. is.

[0118] Next, the affine-coherent prediction motion vector candidate derivation unit 463 performs affine-coherent prediction. The affine identical predictor motion vector candidate derivation unit 463 performs the following process: The same processing as that of the affine identical predictor motion vector candidate derivation unit 363 in the encoding device of the present application is.

[0119] The sub-block predicted motion vector candidate selection unit 467 selects the affine inheritance predicted motion vector candidate. The auxiliary derivation unit 461, the affine construction predicted motion vector candidate derivation unit 462, the affine identical predicted motion vector candidate derivation unit 463, and the affine identical predicted motion vector candidate derivation unit 464 are connected. Among the sub-block predicted motion vector candidates derived by the motion vector candidate derivation unit 463, Then, based on the predicted motion vector index transmitted from the encoding device and decoded, a sub-block predicted motion vector candidate is selected, and the selected sub-block predicted motion vector candidate is selected; The information about the complement is supplied to the motion compensation prediction unit 406 and the addition operation unit 467 .

[0120] The addition operation unit 467 adds the sub-block predicted motion vector candidate selected by the sub-block predicted motion vector candidate selection unit 466. The differential motion vector transmitted from the encoding device and decoded is used as the subblock predicted motion vector. The motion vector generated by the addition is supplied to the motion compensation prediction unit 406 .

[0121] <Affine Inheritance Prediction Motion Vector Candidate Derivation> The affine succession prediction motion vector candidate derivation unit 361 will now be described. The motion vector candidate derivation unit 461 also uses the affine inheritance predicted motion vector candidate derivation unit 36 Same as 1.

[0122] The affine inheritance motion vector predictor candidate inherits the motion vector information of the affine control point. FIG. 42 is a diagram for explaining derivation of affine inheritance predicted motion vector candidates.

[0123] The affine inheritance motion vector predictor candidate is a candidate of the spatially adjacent coded and decoded blocks. The motion vectors are obtained by searching for the affine control points that have the same motion vector.

[0124] Specifically, the block (A0, A1) adjacent to the left of the block to be coded / decoded, From the blocks (B0, B1, B2) adjacent to the upper side of the block to be coded / decoded, A maximum of one affine mode is searched for and used as the affine succession predicted motion vector.

[0125] FIG. 46 is a flowchart of deriving affine inheritance predicted motion vector candidates. First, the block (A0, A1) adjacent to the left of the block to be coded / decoded is treated as a left group. The block containing A0 is a block using affine guarantee (Affine If A0 is in affine mode (S3102), it is determined whether A0 is in affine mode (S 3102:YES), obtain the affine model used by A0 (S3103), and Proceed to processing of adjacent blocks. If A0 is not in affine mode (S3102: NO), The target of deriving affine inheritance predicted motion vector candidates is A0->A1, and the block including A1 Attempt to get affine mode from

[0126] Next, the blocks (B0, B1, B2) adjacent to the upper side of the block to be coded / decoded are It is determined whether the block including B0 is in the affine mode (S3104). If B0 is in affine mode (S3105: YES), The affine model used by B0 is acquired (S3106), and the process ends. If the affine inheritance prediction motion vector candidate is not the target (S3105: NO), B0->B1, and try to get the affine mode from the block containing B1. If 1 is not the affine mode (S3105: NO), the affine inheritance predicted motion vector candidate The target of the supplementary derivation is B1->B2, and an attempt is made to obtain an affine mode from the block including B2. do.

[0127] In this way, we divided the group into left and upper blocks, and for the left block, searches for affine models in the order from the bottom left to the top left block, and for the left block, By searching the affine model in order from the top right to the top left block, we can find two as different as possible. The affine model can be obtained by using the differential motion vectors. Therefore, it is possible to derive affine predicted motion vector candidates that have a small motion vector.

[0128] <Affine Construction Prediction Motion Vector Candidate Derivation> The affine construction prediction motion vector candidate derivation unit 362 will now be described. The motion vector candidate derivation unit 462 is also the same as the affine construction prediction motion vector candidate derivation unit 36 Same as 2.

[0129] The affine constructed motion vector candidate is derived from the motion information of spatially adjacent blocks. The motion vector information of the control points is constructed.

[0130] FIG. 43 is a diagram illustrating the derivation of affine-construction predicted motion vector candidates. The affine construction motion vector predictor candidate is a candidate of the spatially adjacent coded and decoded blocks. The motion vectors are combined to construct a new affine model. Specifically, the blocks adjacent to the upper left of the block to be coded / decoded (B2, B3, A2 ) to derive the motion vector of the upper left affine control point CP0, and The motion vector from the upper right adjacent block (B1, B0) to the upper right affine control point CP1 and then calculate the value from the block (A1, A0) adjacent to the bottom left of the block to be coded / decoded. The motion vector of the lower affine control point CP2 is derived.

[0131] FIG. 47 is a flowchart showing the derivation of affine constructed motion vector predictor candidates. First, the upper left control point CP0, the upper right control point CP1, and the lower left affine control point CP2 are derived ( The upper left affine control point CP0 is the same as the reference image of the block to be coded or decoded. Calculated by searching for a reference block with an image in the order of B2, B3, and A2 reference blocks. The upper right affine control point CP1 has the same reference image as the block to be coded or decoded. The reference blocks are calculated by searching the B1, B0 reference blocks in that order. The affine control point CP2 is a reference block that has the same reference image as the block to be coded or decoded. The calculation is performed by searching the A1, A0 reference blocks in that order.

[0132] When selecting the three affine control points mode as the affine construction predicted motion vector (S 3202:YES), all three affine control points (CP0, CP1, CP2) are derived. It is determined whether the three affine control points (CP0, CP1, CP2) ) are all derived (S3203: YES), the three affine control points (CP0,C P1, CP2) is used as the affine model for the affine constructed predicted motion vector (S32 04). If you do not select the three affine control points mode and select the two affine control points mode, If (S3202:NO), then both affine control points (CP0, CP1) are derived. It is determined whether or not the two affine control points (CP0, CP1) are all derived (S3205). If this is requested (S3205:YES), two affine control points (CP0, CP1) are used. The obtained affine model is used as the affine constructed predicted motion vector (S3206).

[0133] <Derivation of affine identical predictive motion vector candidates> The affine coherent prediction motion vector candidate derivation unit 363 will be described. The motion vector candidate derivation unit 463 also uses the same affine identical predictive motion vector candidate derivation unit 36 Same as 3.

[0134] The affine identical predicted motion vector candidate derives the same motion vector at each affine control point. This can be obtained by doing the following.

[0135] Specifically, similarly to the affine construction motion vector candidate derivation units 362 and 462, each Derive fin control point information and set all affine control points to the same value between CP0 and CP2. In addition, the temporal motion vector derived in the same way as in the normal predicted motion vector mode is can also be obtained by setting the sine vector to all affine control points.

[0136] <Sub-block merge mode derivation> Sub-block merge mode derivation will now be described.

[0137] FIG. 40 shows the block diagram of the sub-block merge mode derivation unit 304 in the encoding device of the present application. The sub-block merge mode derivation unit 304 generates a sub-block merge candidate list This is the same as in the normal merge mode derivation unit 302. This is similar to the merge candidate list mergeCandList, and it is a list of candidates that differs for each subblock. The only difference is that it becomes a strike.

[0138] First, the sub-block temporal merge candidate derivation unit 381 determines the sub-block temporal merge candidates. The details of the derivation of sub-block temporal merge candidates will be described later.

[0139] Next, the affine inheritance merge candidate derivation unit 382 derives affine inheritance merge candidates. The details of deriving affine inheritance merge candidates will be described later.

[0140] Next, the affine construction merge candidate derivation unit 383 derives affine construction merge candidates. The details of deriving affine construction merge candidates will be described later.

[0141] Next, the affine-fixed merge candidate derivation unit 384 derives affine-fixed merge candidates. The details of deriving affine-fixed merge candidates will be described later.

[0142] The sub-block merge candidate selection unit 386 includes the sub-block temporal merge candidate derivation unit 381, An affine inheritance merge candidate derivation unit 382, ​​an affine construction merge candidate derivation unit 383, an affine Among the sub-block merging candidates derived by the fixed merging candidate derivation unit 384, Select a block merge candidate and view information about the selected sub-block merge candidate. The prediction mode determination unit 305 receives the result.

[0143] FIG. 41 shows the block diagram of the sub-block merge mode derivation unit 404 in the decoding device of the present application. The sub-block merging mode derivation unit 404 generates a sub-block merging candidate list su bblockMergeCandList, which is a sub-block merge mode derivation unit 304 and They are the same thing.

[0144] First, the sub-block temporal merge candidate derivation unit 481 determines the sub-block temporal merge candidates. The process of the sub-block time merge candidate derivation unit 481 is to derive the sub-block time marker. This is the same as the processing of the candidate derivation unit 381.

[0145] Next, the affine inheritance merge candidate derivation unit 482 derives affine inheritance merge candidates. The processing of the affine inheritance merge candidate derivation unit 482 is the same as that of the affine inheritance merge candidate derivation unit 3 This is the same as the process in 82.

[0146] Next, the affine construction merge candidate derivation unit 483 derives affine construction merge candidates. The processing of the affine construction merge candidate derivation unit 483 is the same as that of the affine construction merge candidate derivation unit 3 This is the same as the process in 83.

[0147] Next, the affine-fixed merge candidate derivation unit 485 derives affine-fixed merge candidates. The processing of the affine fixed merge candidate derivation unit 485 is the same as that of the affine fixed merge candidate derivation unit 485. This is the same as the process in 85.

[0148] The sub-block merge candidate selection unit 486 includes the sub-block temporal merge candidate derivation unit 481, An affine inheritance merge candidate derivation unit 482, an affine construction merge candidate derivation unit 483, an affine Among the sub-block merging candidates derived by the fixed merging candidate derivation unit 484, the code Selecting sub-block merging candidates based on the index transmitted and decoded from the decoder and provides information about the selected sub-block merge candidates to the motion compensation prediction unit 406. .

[0149] <Sub-block temporal merge candidate derivation> The operation of the sub-block temporal merge candidate derivation unit 381 will be described later.

[0150] <Affine inheritance merge candidate derivation> The affine inheritance merge candidate derivation unit 382 will be described. The affine inheritance merge candidate derivation unit 482 is similar to the affine inheritance merge candidate derivation unit 382 .

[0151] Affine inheritance merge candidates are derived from the affine models of spatially adjacent blocks. The affine model of the fin control points is inherited. It is determined by the size of the block and the motion information of the affine control points.

[0152] FIG. 44 is a diagram for explaining the derivation of affine inheritance merge candidates. The derivation of the motion vector candidates is similar to the derivation of the affine inheritance predicted motion vector, but it is based on spatially adjacent motion vectors. The motion vectors are obtained by searching for the affine control points of the coded and decoded blocks. can be done.

[0153] Specifically, the block (A0, A1) adjacent to the left of the block to be coded / decoded, From the blocks (B0, B1, B2) adjacent to the upper side of the block to be coded / decoded, At most one affine mode is searched for and used for the affine merge mode.

[0154] FIG. 48 is a flowchart of affine inheritance merge candidate derivation. First, the block (A0, A1) adjacent to the left of the block to be coded / decoded is treated as a left group. It is then determined whether the block containing A0 is in affine mode (S3301). S3302). If A0 is in affine mode (S3102: YES), The affine model is obtained (S3303), and the process moves to the adjacent block above. If is not in affine mode (S3302: NO), it is the target of deriving affine inheritance merge candidates. Set A0->A1 and try to get the affine mode from the block containing A1.

[0155] Next, the blocks (B0, B1, B2) adjacent to the upper side of the block to be coded / decoded are It is then determined whether the block containing B0 is in the affine mode (S3304). If B0 is in affine mode (S3305: YES), The affine model used by B0 is acquired (S3306), and the process ends. If it is not (S3305: NO), the target of affine inheritance merge candidate derivation is B0->B. 1 and try to get the affine mode from the block containing B1. If the mode is not affine inheritance merge candidate derivation, the target of affine inheritance merge candidate derivation is B1- >B2 and try to get the affine mode from the block containing B2.

[0156] <Affine Construction Merge Candidate Derivation> The affine construction merge candidate derivation unit 383 will be described. The affine construction merge candidate derivation unit 483 is similar to the affine construction merge candidate derivation unit 383 .

[0157] 45 is a diagram for explaining the derivation of affine construction merge candidates. The motion information of spatially adjacent blocks and the temporal coding block are used to generate the affine image. Construct an affine model of the points.

[0158] Specifically, the blocks adjacent to the upper left of the block to be coded / decoded (B2, B3, A 2) to derive the motion vector of the upper left affine control point CP0, and The motion vector of the upper right affine control point CP1 is calculated from the block (B1, B0) adjacent to the upper right side of the block. The code is derived from the block (A1, A0) adjacent to the lower left of the block to be coded / decoded. The motion vector of the lower-left affine control point CP2 is derived, and the lower-right side of the block to be coded / decoded is The motion vector of the lower right affine control point CP3 from the adjacent temporal coding block (T0) is Derive.

[0159] FIG. 49 is a flowchart of affine construction merge candidate derivation. First, the upper left affine control point CP0, the upper right affine control point CP1, and the lower left affine control point C P2, the lower right affine control point CP3 is derived (S3401). The upper left affine control point CP0 is obtained by searching for blocks with motion information in the order of B2, B3, and A2 blocks. The upper right affine control point CP1 is used to calculate the motion information of the blocks B1 and B0. The lower left affine control point CP2 is calculated by searching in the order of lock priority. The calculation is performed by searching for the blocks with the same priority, A1, A0. The control point CP3 is calculated by searching the motion information of the time block.

[0160] Next, the derived upper left affine control point CP0, upper right affine control point CP1, and lower left affine control point CP2 are Is it possible to construct an affine model using three affine control points using the affine control point CP2? If it is possible to construct it (S3402: YES), it is called an upper-left affine Three affine control points: CP0, CP1, and CP2. The control point affine model is set as an affine merge candidate (S3403).

[0161] Next, the derived upper left affine control point CP0, upper right affine control point CP1, and lower right affine control point CP2 are Is it possible to construct an affine model using three affine control points using the affine control point CP3? If it is possible to construct it (S3404: YES), it is called an upper-left affine Three affine control points: CP0, CP1, and CP3. The control point affine model is set as an affine merge candidate (S3405).

[0162] Next, the derived upper left affine control point CP0, lower left affine control point CP2, and lower right affine control point CP3 are Is it possible to construct an affine model using three affine control points using the affine control point CP3? If it is possible to construct it (S3406: YES), it is called an upper-left affine Three affine control points: CP0, CP2, and CP3. The control point affine model is set as an affine merge candidate (S3407).

[0163] Next, two lines are drawn using the derived upper left affine control point CP0 and upper right affine control point CP1. It is determined whether or not an affine model can be constructed using affine control points (S3408). If construction is possible (S3408: YES), the upper-left affine control point CP0, the upper-right affine The two-affine control point affine model with control point CP1 is set as an affine merge candidate (S 3409).

[0164] Next, two lines are drawn using the derived upper left affine control point CP0 and lower left affine control point CP2. It is determined whether an affine model can be constructed using affine control points (S3410). If construction is possible (S3410: YES), the upper left affine control point CP0, the lower left affine The two-affine control point affine model with control point CP2 is the affine merge candidate (S 3411).

[0165] Here, whether or not an affine model is constructed is determined based on the following conditions. 1. The reference image for all affine control points is the same (affine transformation is possible). 2. At least one affine control point has a different motion vector (can be expressed as a translation). (I can't come)

[0166] In this way, in the first embodiment, the upper left affine constructor is used for all affine construct merge candidates. The top-left affine control point CP0 is the control point from left to right and from top to bottom of the image. When processing in the raster scan order below, there is a possibility that coded and decoded blocks exist. Highest, i.e., most likely to have motion information present.

[0167] The upper left affine control point CP0 is used to derive affine construction merge candidates. The amount of processing required to determine whether the reference images are the same when combining control points is reduced. In addition, the upper right affine control point CP1, the lower left affine control point CP2, and the lower right affine control point C In the derivation of the affine control point of P3 (S3401), the upper left affine control point is the same as CP0. In this case, it is possible to search for the same reference image. Affine construction merge candidates are derived, and valid affine construction merge candidates with the same reference image are selected. Therefore, affine merge candidates with high coding efficiency can be derived.

[0168] Here, we will explain in more detail how to determine the reference image based on the upper left affine control point CP0. explain.

[0169] Figure 62 shows the combination of each derived affine control point in the affine construction merge candidate derivation. This figure shows the comparison target of motion information when deriving affine construction merge candidates by combining them. be.

[0170] FIG. 62(A) shows the upper left affine control point CP0, the upper right affine control point CP1, and the lower left affine control point CP2. 1 is a diagram showing a case where a three-affine control point affine model is constructed using the affine control point CP2. In this case, the motion information of the upper left affine control point CP0 and the upper right affine control point CP1 is compared, Whether the reference images of the upper left affine control point CP0 and the upper right affine control point CP1 are the same, and The difference between the motion vectors of the upper left affine control point CP0 and the upper right affine control point CP1 is compared. Similarly, the motion information of the upper left affine control point CP0 and the lower left affine control point CP2 is To be compared.

[0171] FIG. 62(B) shows the upper left affine control point CP0, the lower left affine control point CP2, and the lower right affine control point CP3. 1 is a diagram showing a case where a three-affine control point affine model is constructed using the affine control point CP3. In this case, the motion information of the upper left affine control point CP0 and the lower left affine control point CP2 is compared, The motion information of the upper left affine control point CP0 and the lower right affine control point CP3 is compared.

[0172] Figure 62(C) shows the upper left affine control point CP0, the upper right affine control point CP1, and the lower right affine control point CP2. 1 is a diagram showing a case where a three-affine control point affine model is constructed using the affine control point CP3. In this case, the motion information of the upper left affine control point CP0 and the upper right affine control point CP1 is compared, The motion information of the upper left affine control point CP0 and the lower right affine control point CP3 is compared.

[0173] Figure 62(D) shows a two-axis affine control point with the upper left affine control point CP0 and the upper right affine control point CP1. This is a diagram showing the construction of a fin control point affine model. In this case, the upper left affine control point The motion information of CP0 and the upper right affine control point CP1 is compared.

[0174] Figure 62(E) shows a two-axis affine control point with an upper left affine control point CP0 and a lower left affine control point CP2. This is a diagram showing the construction of a fin control point affine model. In this case, the upper left affine control point The motion information of CP0 and the bottom-left affine control point CP2 is compared.

[0175] In this way, in all combinations of affine construction merge candidate derivations, the upper-left affine The motion information of the control point CP0 and each affine control point CPx (x=1, 2, 3) is compared, It is determined whether the reference images are identical and affine transformation is possible.

[0176] <Affine-fixed merge candidate derivation> The affine-fixed merge candidate derivation unit 385 will be described. The affine fixed merge candidate derivation unit 485 is similar to the affine fixed merge candidate derivation unit 385. The candidate page fixes the motion information of the affine control points with the fixed motion information. The motion vector of each affine control point is fixed to (0,0).

[0177] <Temporal motion vector prediction> Before explaining the temporal motion vector prediction, the temporal relationship between pictures will be explained. FIG. 55(a) shows a case where the current coding block and the current coding picture are different in time. This shows the relationship between the coded pictures. A specific coded picture is defined as ColPic. ColPic is specified by the syntax. can be.

[0178] Also, in FIG. 55(b), in ColPic, the same position as the coding block to be coded and These coding blocks T 0 and T1 are at almost the same position in a picture that is temporally different from the picture to be coded. is a coding block of

[0179] The above explanation of the temporal relationship of pictures is for encoding, but the same applies to decoding. In other words, when decoding, the encoding in the above explanation can be replaced with decoding and the explanation can be repeated in the same way. It will be revealed.

[0180] The temporal motion vector predictor candidate derivation in the normal motion vector predictor mode derivation unit 301 in FIG. The operation of the output unit 322 will be described with reference to FIG.

[0181] First, ColPic is derived (step S4201). For the derivation of ColPic, see FIG. This will be explained in light of the above.

[0182] If the slice type slice_type is a B slice and the flag collocated_from_l0_flag is 0, If (step S4211: YES, step S4212: YES), RefPicList1[0], That is, the picture with reference index 0 in the reference list L1 is a picture colPic at a different time. If not, that is, if the slice type slice_type is B slice, If the flag collocated_from_l0_flag is 1 (step S4211: YE S, step S4212: NO), or if the slice type slice_type is a P slice (Step S4211: NO, Step S4214: YES), RefPicList0[0], that is, That is, the picture with reference index 0 in reference list L0 is the picture colPic at a different time. If slice_type is not a P slice (step S4214: If NO, the process ends.

[0183] Again, refer to Figure 56. After ColPic is derived, the coding block colCb is derived and the code This process will be described with reference to FIG. do.

[0184] First, in a picture colPic at a different time, the bottom right ( The coding block located outside is defined as a coding block colCb at a different time (step S4221) This coding block corresponds to coding block T0 in FIG.

[0185] Next, the coding information of the coding block colCb at a different time is obtained (step S422 2) If the PredMode of the coding block colCb of a different time is not available or the coding block colCb of a different time is When the prediction mode PredMode of the coded block colCb is intra prediction (MODE_INTRA) ( Step S4223: NO, Step S4224: YES), in the picture colPic at different times The coding block located at the bottom right of the center of the same position as the coding block to be processed is processed at a different time. The coding block between the two is designated as colCb (step S4225). This corresponds to the coded block T1.

[0186] Again, see Figure 56. Next, for each reference list, inter prediction information is derived (S Here, for the coding block colCb, the motion of each reference list is Derive the vector mvLXCol and the flag availableFlagLXCol that indicates whether the coding information is valid. LX indicates the reference list, and in the derivation of reference list 0, LX becomes L0, and in the derivation of reference list 1, Then, LX becomes L1. The derivation of inter prediction information will be described with reference to FIG.

[0187] If a different time coding block colCb is not available (S4231S4231:NO) Or, if the prediction mode PredMode is intra prediction (MODE_INTRA) (S4232: NO) The flag availableFlagLXCol and the flag predFlagLXCol are both set to 0 (step S4233 ), and the motion vector mvLXCol is set to (0,0) (S4234), and the process ends.

[0188] The coding block colCb is available (S4231: YES) and the prediction mode PredMode is If it is not a tra prediction (MODE_INTRA) (S4232:YES), use the following procedure to set mvCol and refIdx Calculate Col and availableFlagCol.

[0189] Flag PredFlagL0[xP If [Col][yPCol] is 0 (S4235: YES), the prediction mode of the coding block colCb is Pr ed_L1, the motion vector mvCol is the L1 motion vector of the coding block colCb. The reference index refIdxCol is set to the same value as the MvL1[xPCol][yPCol] (S4236). is set to the same value as the L1 reference index RefIdxL1[xPCol][yPCol] (S4237), The list ListCol is set to L1 (S4238). Here, xPCol and yPCol are different time is an index indicating the position of the top left pixel of the coding block colCb in the picture colPic. be.

[0190] On the other hand, if the L0 prediction flag PredFlagL0[xPCol][yPCol] of the coding block colCb is not 0, If the result is NO (S4235), the L1 prediction flag PredFlagL1[xPCol][yPC ol] is 0. The L1 prediction flag PredFlagL1[xPCol][ yPCol is 0 (S4239: YES), the motion vector mvCol is the It is set to the same value as the L0 motion vector MvL0[xPCol][yPCol] (S4240), The index refIdxCol is set to the same value as the reference index RefIdxL0[xPCol][yPCol] of L0. The list ListCol is set to L0 (S4242).

[0191] L0 prediction flag PredFlagL0[xPCol][yPCol] of coding block colCb and coding block col If both Cb L1 prediction flags PredFlagL1[xPCol][yPCol] are not 0 (S4235: NO, S4239:NO), and the inter prediction mode of the coding block colCb is bi-prediction (Pred_BI). Therefore, one of the two motion vectors L0 and L1 is selected (S4243). 0 is the coding block when the inter prediction mode of the coding block colCb is bi-prediction (Pred_BI). 10 is a flowchart illustrating a procedure for deriving inter prediction information of a lock.

[0192] First, the POCs of all pictures registered in all reference lists are checked against the current encoding. It is determined whether the POC is smaller than the POC of the target picture (S4251). POC of all pictures registered in L0 and L1, which are all reference lists of b is smaller than the POC of the current picture to be coded (S4251: YES), and LX is L0, that is, the predicted vector candidate of the motion vector of L0 of the coding block to be coded is derived. If yes (S4252: YES), inter prediction for L0 of coding block colCb is performed. LX is the predicted motion vector of L1 of the coding block to be coded. If the measurement vector candidate is derived (S4252: NO), the L1 On the other hand, the inter prediction information of the coding block colCb is selected. At least one of the POCs of the pictures registered in L0 and L1 is the current picture to be coded. If the POC is larger than the current one (S4251:NO), the flag collocated_from_l0_fla If g is 0 (S4253: YES), inter prediction information for L0 of the coding block colCb If the flag collocated_from_l0_flag is 1 (S4253: NO), The inter prediction information for L1 of the locked colCb is selected.

[0193] When selecting inter prediction information for L0 of coding block colCb (S4252:Y ES, S4253:YES), the motion vector mvCol is set to the same value as MvL0[xPCol][yPCol] (S4254), and the reference index refIdxCol becomes the same value as RefIdxL0[xPCol][yPCol]. The list ListCol is set to L0 (S4256).

[0194] When selecting inter prediction information for L1 of coding block colCb (S4252:N O, S4253:NO), the motion vector mvCol is set to the same value as MvL1[xPCol][yPCol]. (S4257), the reference index refIdxCol is set to the same value as RefIdxL1[xPCol][yPCol]. (S4258), and the list ListCol is set to L1 (S4259).

[0195] Returning to Figure 59, once the inter prediction information is obtained from the coding block colCb, the flag ava Both ilableFlagLXCol and flag predFlagLXCol are set to 1 (S4244).

[0196] Next, the motion vector mvCol is scaled to obtain a motion vector mvLXCol (S42 45S4245). The scaling calculation procedure for this motion vector mvLXCol is shown in Figure 61. and explain.

[0197] The POC of the picture colPic at different times is referenced by a list ListCol of coding blocks colCb. The POC of the reference picture corresponding to the reference index refIdxCol is subtracted to obtain the picture. The distance td between the two pictures is calculated (S4261). If the POC of the reference picture referenced in the list ListCol of block colCb comes before the POC of the reference picture in the display order, In this case, the inter-picture distance td is a positive value, and the coding block is closer to the picture colPic at a different time. If the POC of the reference picture referenced in the list ListCol of the check colCb is later in the display order, The inter-picture distance td is a negative value. td = POC of pictures colPic at different times - ListCol of coding blocks colCb POC of the reference picture to be referenced in

[0198] The list LX of pictures currently to be coded is referenced from the POC of the picture currently to be coded. The POC of the reference picture is subtracted from the POC of the reference picture to calculate the inter-picture distance tb (S4262). The list LX of the picture to be coded is referred to rather than the picture to be coded. If the reference picture is earlier in display order, the inter-picture distance tb is a positive value, and the current code If the reference picture in the list of pictures to be coded LX is later in display order, The inter-cha distance tb is a negative value.

[0199] tb = POC of current picture to be coded / decoded - LX reference image of temporal merge candidate POC of the reference picture corresponding to the index

[0200] Next, the inter-picture distances td and tb are compared (S4263). If b is equal (S4263: YES), the motion vector mvLXCol is calculated using the following formula ( S4264), this scaling calculation process ends.

[0201] mvLXCol = mvCol

[0202] On the other hand, if the inter-picture distances td and tb are not equal (S4263: NO), the following formula is used: The variable tx is calculated (S4265).

[0203] tx = ( 16384 + Abs( td ) >> 1 ) / td

[0204] Next, the scaling coefficient distScaleFactor is calculated using the following equation (S4266). distScaleFactor = Clip3( -4096, 4095, ( tb * tx + 32 ) >> 6 ) Here, Clip3(x,y,z) is a function that limits the value z to a minimum value of x and a maximum value of y. Then, the motion vector mvLXCol is calculated by the following formula (S4267), and this scaling calculation is End processing.

[0205] mvLXCol = Clip3( -32768, 32767, Sign( distScaleFactor * mvLXCol ) * ( (Abs( distScaleFactor * mvLXCol ) + 127 ) >> 8 ) )

[0206] Here, Sign(x) is a function that returns the sign of the value x, and Abs(x) is a function that returns the absolute value of the value x. .

[0207] Again, refer to FIG. 56. Then, the motion vector mvL0Col of L0 is calculated by the normal prediction motion vector mvL0Col. The motion vector prediction candidate list mvpListLXN in the motion vector mode derivation unit 301 is a candidate. However, this addition is performed for the coded block colCb of the reference list 0. This is only possible when the flag availableFlagL0Col=1 indicates whether the L1 motion vector is available. The vector mvL1Col is calculated by the predicted motion vector mode derivation unit 301. The candidate is added to the list of candidates mvpListLXN (S4205). Flag indicating whether the coding block colCb of reference list 1 is valid or not: availableFlagL1Col=1 This is the only step. The process of the temporal motion vector predictor candidate derivation unit 322 is then completed. The above description of the normal predicted motion vector mode derivation unit 301 is for the encoding process. The same applies to decoding. That is, in the normal predicted motion vector mode derivation unit 401 in FIG. The operation of the temporal motion vector predictor candidate derivation unit 422 is the same as that of the encoding in the above description. Substituted, they are explained similarly.

[0208] <Time Merge> The operation of the temporal merge candidate derivation unit 342 in the normal merge mode derivation unit 302 in FIG. This will be described with reference to FIG.

[0209] First, ColPic is derived (step S4301). Next, the coding block colCb is derived. Then, the encoding information is acquired (step S4302). The above process derives the temporal prediction motion vector. Since these steps are the same as steps S4201 to S4204 in the rule candidate derivation unit 322, their explanations will be omitted. do.

[0210] Next, a flag availableFlagCol indicating whether the coding block colCb is valid or not is calculated ( S4305). If the flag availableFlagL0Col or the flag availableFlagL1Col is 1 availableFlagCol will be 1 otherwise availableFlagCol will be 0.

[0211] Then, the motion vector mvL0Col of L0 and the motion vector mvL1Col of L1 are calculated by the normal motion vector calculation method described above. The merge candidate list mergeCandList in the page mode derivation unit 302 is added as a candidate. (S4306). However, this addition is performed by setting a flag a indicating whether the coding block colCb is valid or not. This is only the case when validFlagCol=1. Finish.

[0212] The above explanation of the temporal merge candidate derivation unit 342 is for encoding, but the same applies to decoding. That is, the temporal merge candidate derivation in the normal merge mode derivation unit 402 in FIG. The operation of unit 442 is similarly described, substituting decoding for encoding in the above description.

[0213] <Update of historical motion vector predictor candidate list> Next, the encoding information storage memory 111 on the encoding side and the encoding information storage memory 20 on the decoding side are Methods for initializing and updating the historical motion vector predictor candidate list HmvpCandList in preparation for 5 FIG. 26 shows the procedure for initializing and updating the history motion vector predictor candidate list. 1 is a flowchart illustrating the above.

[0214] In this embodiment, the historical motion vector predictor candidate list HmvpCandList is updated based on the coding information. This is implemented in the information storage memory 111 and the coded information storage memory 205. The prediction unit 102 and the inter-prediction unit 203 include a history prediction motion vector candidate list update unit. It may be set to update the historical motion vector predictor candidate list HmvpCandList.

[0215] At the beginning of the slice, the historical motion vector prediction candidate list HmvpCandList is initialized. On the encoding side, the prediction method decision unit 105 decides whether to use the normal predicted motion vector mode or the normal merge mode. When the selected motion vector prediction candidate list HmvpCandList is updated, The prediction information decoded by the bitstream decoding unit 201 is in the normal predicted motion vector mode or the normal In the merge mode, the historical motion vector predictor candidate list HmvpCandList is updated.

[0216] Used when performing inter prediction in normal predicted motion vector mode or normal merge mode. The inter prediction information is used as the inter prediction information candidate hMvpCand in the history prediction motion vector candidate list. The inter prediction information candidate hMvpCand is registered in the HmvpCandList. The reference index for L0 and L1, refIdxL1, indicates whether L0 prediction is performed. an L0 prediction flag predFlagL0 and an L1 prediction flag predFl indicating whether or not L1 prediction is performed; It includes agL1, a motion vector mvL0 for L0, and a motion vector mvL1 for L1.

[0217] The encoding information storage memory 111 on the encoding side and the encoding information storage memory 205 on the decoding side are provided. The elements (i.e., the image) registered in the history motion vector predictor candidate list HmvpCandList are Among the inter prediction information candidates, there is inter prediction information with the same value as the inter prediction information candidate hMvpCand. If it exists, remove the element from the historical motion vector predictor candidate list HmvpCandList. On the other hand, if there is no inter-prediction information with the same value as the inter-prediction information candidate hMvpCand, , delete the first element of the history predicted motion vector candidate list HmvpCandList, and The inter-prediction information candidate hMvpCand is added to the end of the vector candidate list HmvpCandList.

[0218] The encoding information storage memory 111 on the encoding side and the encoding information storage memory 2 on the decoding side according to the present invention The number of elements in the historical motion vector predictor candidate list HmvpCandList for 05 is six.

[0219] First, the historical motion vector predictor candidate list HmvpCandList is initialized for each slice. (Step S2101 in FIG. 26) At the beginning of the slice, the historical predicted motion vector candidate list Hm All elements of vpCandList are emptied and registered in the history predicted motion vector candidate list HmvpCandList. The number of recorded historical motion vector predictor candidates (current number of candidates). The value of NumHmvpCand is set to 0. do.

[0220] Note that the initialization of the historical motion vector predictor candidate list HmvpCandList is performed on a slice-by-slice basis ( However, it is possible to perform this on a picture-by-picture, tile-by-tile or tree block basis. It may be performed on a row-by-row basis.

[0221] Next, for each coding block in the slice, the following historical motion vector predictor candidate list Hmvp The CandList update process is repeated (steps S2102 to S2107 in FIG. 26).

[0222] First, the initial setting is performed for each coding block. Set the value of the parameter "identicalCandExist" to FALSE (false) and delete the target index that indicates the candidate for deletion. The index removeIdx is set to 0 (step S2103 in FIG. 26).

[0223] It is determined whether or not there is an inter-prediction information candidate hMvpCand to be registered (steps in FIG. 26 ). The prediction method determination unit 105 on the encoding side determines whether the normal prediction motion vector mode or When the normal merge mode is determined, or when the bitstream decoding unit 201 on the decoding side performs normal prediction, When decoded as a motion vector mode or normal merge mode, the inter prediction information The information is set as the inter prediction information candidate hMvpCand to be registered. Intra prediction mode, sub-block prediction motion vector mode or sub-block merge mode, or when the bitstream decoding unit 201 on the decoding side determines that the bitstream is an intra-prediction mode, Decoded as sub-block predicted motion vector mode or sub-block merge mode In this case, the historical motion vector predictor candidate list HmvpCandList is not updated, and the registered image is There is no inter prediction information candidate hMvpCand. If it does not exist, steps S2105 to S2106 are skipped (step S in FIG. 26). 2104: NO). If there is an inter-prediction information candidate hMvpCand to be registered, The processing from S2105 onwards is carried out (step S2104 in FIG. 26: YES).

[0224] Next, the input to be registered is added to each element of the history motion vector predictor candidate list HmvpCandList. The elements (inter prediction information) with the same value as the inter prediction information candidate hMvpCand, i.e., the same elements It is determined whether or not it exists (step S2105 in FIG. 26). 10 is a flowchart of a processing procedure. (Step S2121 in FIG. 27: NO), the historical motion vector predictor candidate list HmvpCa Since the ndList is empty and there are no identical candidates, steps S2122 to S2125 in FIG. 27 are executed. The number of historical motion vector predictor candidates NumHmvpC is If the value of and is greater than 0 (step S2121 in FIG. 27: YES), the historical predicted motion vector The process of step S2123 is repeated for the data index hMvpIdx from 0 to NumHmvpCand-1. First, the history of the motion vector predictor candidate list is returned (steps S2122 to S2125 in FIG. 27). The hMvpIdx-th element HmvpCandList[hMvpIdx] counting from 0 in the list is the inter prediction information candidate hM It is compared with vpCand to see if it is the same (step S2123 in FIG. 27). Step S2123: YES), a flag "identicalCandE" indicating whether an identical candidate exists or not, Set the value of xist to TRUE and the deletion target index remo to indicate the position of the element to be deleted. Set the current value of the history predicted motion vector index hMvpIdx to veIdx and check the same element. If they are not the same (step S2123 in FIG. 27: NO), increase hMvpIdx by 1. If the history predicted motion vector index hMvpIdx is less than or equal to NumHmvpCand-1, If so, the process from step S2123 onwards is carried out.

[0225] Returning to the flowchart of FIG. 26 again, the historical predicted motion vector candidate list HmvpCandList The element shift and addition process is performed (step S2106 in FIG. 26). Element shift / addition processing of the history predicted motion vector candidate list HmvpCandList in step S2106 First, the historical motion vector predictor candidate list HmvpCandList is Either remove the stored elements and then add the new elements, or add the new elements without removing any elements. Specifically, the flag "identicalCandE" indicates whether or not an identical candidate exists. xist is compared to determine whether it is TRUE or NumHmvpCand is 6 (step S2141 in FIG. 28). ) The flag identicalCandExist indicates whether an identical candidate exists or not. If the number of candidates NumHmvpCand is 6, any of the conditions is satisfied (step S2141 in FIG. 28: YES), excluding the elements stored in the historical motion vector predictor candidate list HmvpCandList Add a new element from the index i. Set the initial value of the index i to the value of removeIdx + 1. The element shift process in step S2143 is repeated from the initial value of NumHmvpCand to NumHmvpCand. Steps S2142 to S2144 of 28. HmvpCandList[i - 1] is set to HmvpCandList[i] Shift the elements forward by copying the elements of i (step S2143 in FIG. 28), and Increment by 1 (steps S2142 to S2144 in FIG. 28). The (NumHmvpCand-1)th HmvpCandList, counting from 0, corresponds to the end of the motion vector candidate list. The inter prediction information candidate hMvpCand is added to t[NumHmvpCand-1] (step S214 in FIG. 28). 5) Finish the element shift and addition process of this history predicted motion vector candidate list HmvpCandList. On the other hand, the flag identicalCandExist, which indicates whether an identical candidate exists, is set to TRUE. If neither of the conditions of NumHmvpCand is 6 is satisfied (step S2141 in FIG. 28: NO), ), without excluding the elements stored in the history motion vector predictor candidate list HmvpCandList, The inter prediction information candidate hMvpCand is added to the end of the historical prediction motion vector candidate list (see Figure 2). 8, step S2146). Here, the end of the history motion vector predictor candidate list is from 0 to It is the NumHmvpCand-th HmvpCandList[NumHmvpCand] counting from the beginning. Increment the element shift and The addition process is then completed.

[0226] FIG. 31 is a diagram illustrating an example of a process for updating the history motion vector predictor candidate list. The elements (inter prediction information) are registered in the historical predicted motion vector candidate list HmvpCandList When adding a new element to the list, the previous element of the history motion vector predictor candidate list HmvpCandList is added. The new inter-prediction information is compared with the previous inter-prediction information (Fig. 31A). If the value is the same as the third element HMVP2 from the beginning of the vector candidate list HmvpCandList, The element HMVP2 is deleted from the motion vector predictor candidate list HmvpCandList, and the following elements HMVP3 to HM VP5 is shifted forward one by one and stored in the history prediction motion vector candidate list HmvpCandLis A new element is added to the end of t (FIG. 31B), and the history motion vector predictor candidate list HmvpCan The update of the dList is completed (Figure 31C).

[0227] <Historical motion vector predictor candidate derivation process> Next, the history of predicted motion vector candidates in the normal predicted motion vector mode derivation unit 301 on the encoding side is The prediction motion vector mode derivation unit 323 and the normal prediction motion vector mode derivation unit 401 on the decoding side The processing procedure of step S304 in FIG. 20, which is common processing in the driver candidate derivation unit 423, is Method for deriving historical motion vector predictor candidates from historical motion vector predictor candidate list HmvpCandList FIG. 29 is a flowchart illustrating the procedure for deriving a historical motion vector predictor candidate. This is a flow chart.

[0228] The number of current motion vector predictor candidates numCurrMvpCand is the motion vector predictor candidate list mvpLis The maximum number of elements in tLX (here, 2) or more, or the number of history motion vector predictor candidates is NumHm If the value of vpCand is 0 (step S2201 in FIG. 29: NO), step S22 in FIG. The process from S2202 to S2209 is omitted, and the procedure for deriving the historical motion vector predictor candidate is completed. The current number of motion vector predictor candidates numCurrMvpCand is the motion vector predictor candidate list mvpLis If the number of elements of tLX is smaller than 2, and the number of history motion vector predictor candidates NumHmvpCa If the value of nd is greater than 0 (step S2201 in FIG. 29: YES), Process steps S2202 to S2209.

[0229] Next, the index i is 1 to 4 and the number of historical motion vector predictor candidates is numCheckedHMVP The process of steps S2203 to S2208 in FIG. 29 is repeated until the smaller value of Cand is reached. The current number of motion vector predictor candidates, nu, is returned (steps S2202 to S2209 in FIG. 29). When mCurrMvpCand is equal to or greater than 2, which is the maximum number of elements in the motion vector predictor candidate list mvpListLX. (Step S2203 in FIG. 29: NO), Steps S2204 to S2209 in FIG. 29 The process for deriving the historical motion vector predictor candidate is omitted, and the procedure for deriving the current motion vector predictor candidate is terminated. The number of motion vector candidates, numCurrMvpCand, is the maximum number of elements in the motion vector predictor candidate list, mvpListLX. If it is smaller than 2 (step S2203 in FIG. 29: YES), step S2 in FIG. Processes 204 and beyond.

[0230] Next, the process from step S2205 to S2207 is performed when Y is 0 and 1 (L0 and L1). This is performed for each of the current predicted motion vectors (steps S2204 to S2208 in FIG. 29). The number of motion vector predictor candidates, numCurrMvpCand, is the maximum number of elements in the motion vector predictor candidate list, mvpListLX. If the number of the digits is 2 or more (step S2205 in FIG. 29: NO), the process proceeds from step S2206 in FIG. 29. Therefore, the process of S2209 is omitted, and the procedure for deriving the historical motion vector predictor candidate is terminated. The number of current motion vector predictor candidates numCurrMvpCand is the motion vector predictor candidate list mvpListLX If the number of elements is smaller than 2, which is the maximum number of elements (step S2205 in FIG. 29: YES), The process from step S2206 onwards is carried out.

[0231] Next, the motion vector to be coded / decoded is included in the history predicted motion vector candidate list HmvpCandList. The element of the reference index is the same as the reference index refIdxLX of the vector, and the predicted motion vector If the element is different from any element in the vector list mvpListLX (step S2206 in FIG. 29), :YES), the numCurrMvpCand-th element mvpL counting from 0 in the motion vector predictor candidate list istLX[numCurrMvpCand] is the LY of the history predicted motion vector candidate HmvpCandList[NumHmvpCand - i] (Step S2207 in FIG. 29), and the current motion vector predictor candidate Increment the number of historical motion vector predictor candidates, numCurrMvpCand, by 1. In the list, there is a reference index that is the same as the reference index refIdxLX of the motion vector to be coded / decoded. element of the motion vector predictor list mvpListLX, and an element different from any element of the motion vector predictor list mvpListLX is If not (step S2207 in FIG. 29: NO), the additional processing of step S2207 is skipped. Upload.

[0232] The above processing from step S2205 to S2207 in Figure 29 is performed on both L0 and L1. (Steps S2204 to S2208 in FIG. 29). Increment the index i by 1. , the index i is the smaller value of 4 or the number of historical motion vector predictor candidates NumHmvpCand. In the following cases, the process from step S2203 onward is performed again (steps S2202 to S2203 in FIG. 29). S2209).

[0233] <History merge candidate derivation process> Next, the history merge candidate derivation unit 345 of the normal merge mode derivation unit 302 on the encoding side, This is a common process between the normal merge mode derivation unit 402 and the history merge candidate derivation unit 445. 21, which is the processing procedure of step S404, The method for deriving merge candidates will now be described in detail. Figure 30 shows the history merge candidate derivation process. 10 is a flowchart for explaining the process.

[0234] First, initialization is performed (step S2301 in FIG. 30). Set the value of FALSE for each (rrMergeCand -1)th element and set the variable numOrigMergeCand to the current Set numCurrMergeCand to the number of elements currently in the merge candidate list.

[0235] Next, set the initial value of the index hMvpIdx to 1, and then calculate from this initial value to NumHmvpCand. 30. Then, the additional processing from step S2303 to step S2310 in FIG. 30 is repeated (FIG. 3 Steps S2302 to S2311 of step 0. Elements registered in the current merge candidate list If the number of merge candidates is not less than (MaxNumMergeCand-1), the merge Since merge candidates have been added to all elements in the candidate list, the merge candidate derivation process is The process ends (step S2303 in FIG. 30: NO). If the number of elements in the merge candidate list, numCurrMergeCand, is less than or equal to (MaxNumMergeCand-1), Execute the process from step S2304 onwards. Set the value of sameMotion to FALSE (see Figure 30). (Step S2304) Next, the initial value of index i is set to 0, and The processing of steps S2306 and S2307 in FIG. 30 is repeated up to numOrigMergeCand-1 (FIG. 30 (S2305 to S2308). Counting from 0 in the history motion vector prediction candidate list (NumHmvp Cand - hMvpIdx)-th element HmvpCandList[NumHmvpCand- hMvpIdx] is the merge candidate list Counting from 0, compare whether the value is the same as the i-th element mergeCandList[i] (step in Figure 30). P2306).

[0236] The same value of a merge candidate means that all components (inter prediction mode, reference If the values ​​of the merge candidates (reference index, motion vector) are the same, the merge candidates are considered to have the same value. If the complements are the same and isPruned[i] is FALSE (step S2306 in FIG. 30: YES), Set both sameMotion and isPruned[i] to TRUE (step S2307 in FIG. 30). If the values ​​are not the same (step S2306 in FIG. 30: NO), the process of step S2307 is performed. The process from step S2305 to step S2308 in FIG. 30 is repeated. When the process is completed, compare whether sameMotion is FALSE (step S230 in FIG. 30). 9), if sameMotion is FALSE (step S2309 in Figure 30: YES), that is, That is, the (NumHmvpCand - hMvpIdx)th element counting from 0 in the history motion vector candidate list. Since the element HmvpCandList[NumHmvpCand - hMvpIdx] does not exist in mergeCandList, it is not a merge candidate. The numCurrMergeCand-th mergeCand in the list is mergeCandList[numCurrMergeCand], which stores the history predicted movement vector. The (NumHmvpCand - hMvpIdx)th element of the candidate list, counting from 0, is HmvpCandList[NumHmv pCand - hMvpIdx] and increment numCurrMergeCand by 1 (step 30 in Figure 30). The index hMvpIdx is incremented by 1 (step S23 in FIG. 30). 02), steps S2302 to S2311 in FIG. 30 are repeated. When all elements in the history motion vector prediction candidate list have been checked, or the merge candidate list Once merge candidates have been added to all elements of the record, the process of deriving merge candidates for this history is complete. .

[0237] <Sub-block temporal merge candidate derivation> Sub-block temporal merge candidates in the sub-block merge mode derivation unit 304 in FIG. 16 The operation of the derivation unit 381 will be described with reference to FIG.

[0238] First, it is determined whether the coding block is smaller than 8x8 pixels (S4002). If the coding block is less than 8x8 pixels (S4002: YES), the sub-block time marker The flag indicating the existence of a sub-block candidate is set to availableFlagSbCol=0 (S4003). The process of the temporal merge candidate derivation unit is completed. When vector prediction is disabled or sub-block temporal merging is disabled In this case, the same process as when the coding block is less than 8x8 pixels (S4002: YES) is performed. .

[0239] On the other hand, if the coding block is 8x8 pixels or larger (S4002: NO), the coding picture Next, neighboring motion information of the coding block is derived (S4004).

[0240] The process of deriving adjacent motion information of a coding block will be described with reference to FIG. The process of deriving adjacent motion information is similar to the process of the spatial prediction motion vector candidate derivation unit 321 described above. However, the order of searching adjacent blocks is A0, B0, B1, A1, and B2 is not searched. First, the coding information is acquired for the adjacent block n=A0 (S4052). The information includes a flag, availableFlagN, indicating whether adjacent blocks are available, and a reference list. and the motion vector mvLXN.

[0241] Next, it is determined whether adjacent block n is valid or invalid (S4054). If the flag indicating whether or not the data is available is "availableFlagN=1", the data is valid; otherwise, the data is invalid. If adjacent block n is valid (S4054: YES), the reference index refIdxLXN is The reference index of the adjacent block n is set as refIdxLXn (S4056). The motion vector mvLXN of the adjacent block n is set as the motion vector mvLXn of the adjacent block n (S4056). The process of deriving motion information is completed.

[0242] On the other hand, if the adjacent block n is invalid (S4106: NO), the adjacent block n=B0 is set, The encoding information is acquired (S4104). The same process is repeated for B1 and A1 in the order of loop. The process of deriving neighboring motion information loops until the neighboring blocks are valid, and all If the adjacent blocks A0, B0, B1, and A1 are invalid, the process of deriving the adjacent motion information of the blocks is terminated. Complete.

[0243] Referring again to FIG. 50, after the adjacent motion information is derived (S4004), the temporal motion A vector is derived (S4006). The process of deriving a temporal motion vector will be described with reference to FIG. The temporal motion vector is initialized as tempMv=(0,0) (S4062).

[0244] Next, it is determined whether the adjacent motion information is valid or invalid (S4064). If the flag indicating whether the neighboring movement is available is "availableFlagN=1", it is valid, otherwise it is invalid. If the information is invalid (S4064: NO), the process of deriving the temporal motion vector is terminated. Complete.

[0245] On the other hand, if the neighboring motion information is valid (S4064: YES), the L1 prediction is performed for the neighboring block N. It is determined whether the flag predFlagL1N indicating whether measurement is being used is 1 (S4066). If predFlagL1N=0 (S4066: NO), proceed to the next process (S4078). If L1N=1 (S4066:YES), all the pictures registered in all the reference lists are It is determined whether the POC of the current picture is equal to or less than the POC of the current picture to be coded (S4068 If this determination is true (S4068: YES), the process proceeds to the next step (S4070).

[0246] If the slice type slice_type is a B slice and the flag collocated_from_l0_flag is 0, If (S4070: YES and S4072: YES), ColPic and reference picture RefPicList1[ refIdxL1N] (picture with reference index refIdxL1N in reference list L1) is the same or not. If this determination is true (S4074: YES), the temporal motion Let the vector tempMv = mvL1N (S4076). If this determination is false (S4074: NO), The process proceeds to step S4078. If the slice type slice_type is not a B slice and the flag co If llocated_from_l0_flag is not 0 (S4070:NO or S4072:NO), Proceed to the next process (S4078).

[0247] A flag predFlagL0N indicates whether L0 prediction is used in the neighboring block N. It is determined whether predFlagL0N is 1 (S4078). If predFlagL0N=1 (S4078: YES), Co lPic and reference picture RefPicList0[refIdxL0N] (reference index refIdx of reference list L0) If the picture in L0N is the same as the picture in L1 (S408), the process proceeds to step S409. 80: YES), the temporal motion vector tempMv is set to mvL0N (S4082). If it is false (S4080: NO), the process of deriving the temporal motion vector ends. Again, refer to Figure 50. Next, ColPic is derived (S4016). This process is performed using time prediction. Since this is the same as S4201 in the motion vector candidate derivation unit 322, the explanation will be omitted. .

[0248] Then, a coding block colCb for a different time is set (S4017). The block in the picture ColPic at the same time as the coding block to be processed is located at the bottom right of the center. The coding block is set as colCb. This coding block is the code shown in Figure 55. This corresponds to encryption block T1.

[0249] Next, the position where the temporal motion vector tempMv is added to the coding block colCb is The upper left position of the coding block colCb is (xColCb, yColCb) (S4018). If the temporal motion vector tempMv is (tempMv[0], tempMv[1]) with 1 / 16 pixel accuracy, The top left position of colCb is as follows:

[0250] xColCb = Clip3( xCtb, xCtb + CtbSizeY + 3, xcolCb + ( tempMv[0] >> 4 ) ) yColCb = Clip3( yCtb, yCtb + CtbSizeY - 1, ycolCb + ( tempMv[1] >> 4 ) )

[0251] Here, the top left position of the treeblock is (xCtb, yCtb), and the size of the treeblock is CtbS. As shown in the above formula, the position after adding tempMv is significantly different from before adding tempMv. If this position is outside the screen, it is corrected to a range of about the size of the tree block. If this happens, the image will be corrected within the screen.

[0252] Then, the prediction mode PredMode of this coding block colCb is inter prediction (MODE_INTER If the prediction mode of colCb is not inter prediction (S4020), 4020:NO), flag indicating the existence of sub-block time merge candidates availableFlagSbCol= The value is set to 0 (S4003), and the processing of the sub-block temporal merge candidate derivation unit is completed.

[0253] On the other hand, if the prediction mode of colCb is inter prediction (S4020: YES), Inter prediction information is derived for each of the colCb and colCb (S4022, S4023). , the central motion vector ctrMvLX for each reference list and a flag indicating whether LX prediction is used. Derive the lag ctrPredFlagLX. LX indicates the reference list. In the derivation of reference list 0, LX is L0. In the derivation of reference list 1, LX is L1. Regarding the derivation of inter prediction information, This will be explained with reference to 3.

[0254] If the coding block colCb of a different time is not available (S4112: NO), or if the If the prediction mode PredMode is intra prediction (MODE_INTRA) (S4114: NO), the flag ava The ilableFlagLXCol and flag predFlagLXCol are both set to 0 (step S4116). mvCol is set to (0,0) (S4118), and the process of deriving inter prediction information is terminated. .

[0255] If the coding block colCb is available (S4112: YES) and the prediction mode PredMode is If it is not a tra prediction (MODE_INTRA) (S4114:YES), use the following procedure to set mvCol and refIdx Calculate Col and availableFlagCol.

[0256] Flag PredFlagLX[xP If [Col][yPCol] is 1 (S4120: YES), the motion vector mvCol is the coding block col It is set to the same value as MvLX[xPCol][yPCol], the LX motion vector of Cb (S4122). The reference index refIdxCol is the same as the reference index RefIdxLX[xPCol][yPCol] of LX (S4124), and the list listCol is set to LX (S4126). , xPCol, yPCol are the upper left pixel of the coding block colCb in the picture colPic at different times. is an index indicating the position of

[0257] On the other hand, a flag PredFla indicating whether LX prediction of the coding block colCb is used If gLX[xPCol][yPCol] is 0 (S4120: NO), the following process is performed. First, all The POC of all pictures registered in the reference list is the same as that of the current picture to be coded. It is determined whether the result is below POC (S4128). And LY prediction of colCb is used. It is determined whether the flag PredFlagLY[xPCol][yPCol] indicating whether the current position is 1 or not (S4128). Here, LY prediction is defined as a reference list different from LX prediction. That is, when LX=L0, LY=L1, When LX=L1, LY=L0.

[0258] If this determination is true (S4128: YES), the motion vector mvCol is It is set to the same value as MvLY[xPCol][yPCol], the LY motion vector of Cb (S4130). The reference index refIdxCol is the same as the reference index RefIdxLY[xPCol][yPCol] of LY. (S4132), and the list listCol is set to LX (S4134).

[0259] On the other hand, if this determination is false (S4128: NO), the flag availableFlagLXCol and the flag predFlagLXCol are both set to 0 (step S4116), and the motion vector mvCol is set to (0,0). Then (S4118), the process of deriving inter prediction information ends. If inter prediction information can be obtained from the coding block colCb, the flag availableFlagLXCol and flag predFlagLXCol are both set to 1 (S4136).

[0260] Next, the motion vector mvCol is scaled to obtain a motion vector mvLXCol (S41 38). This process is the same as S4245 in the temporal motion vector predictor candidate derivation unit 322. Therefore, the description will be omitted.

[0261] Referring again to Figure 50, once the inter prediction information is derived for each reference list, the calculated The calculated motion vector mvLXCol is the center motion vector ctrMvLX, and the calculated flag predFlagLXCol is The flag is set to ctrPredFlagLX (S4022, S4023).

[0262] Then, it is determined whether the central motion vector is valid or invalid (S4024). If 0 and ctrPredFlagL1=0, it is invalid, otherwise it is invalid. If it is valid (S4024: NO), the flag avail- able indicates the existence of a sub-block time merge candidate. Set leFlagSbCol=0 (S4003) and execute the process of the sub-block time merge candidate derivation part. Finish.

[0263] On the other hand, if the central motion vector is valid (S4024: YES), the sub-block time marker The flag indicating the existence of a sub-block candidate is set to availableFlagSbCol=1 (S4025). This process will be described with reference to FIG.

[0264] First, the number of sub-blocks in the width direction is calculated from the width cbWidth and height cBheight of the coding block colCb. numSbX and the number of sub-blocks in the height direction numSbY are calculated (S4152). LXSbCol=0 (S4152). After this process, the process is repeated in units of the prediction sub-block colSb. This iteration repeats the process of changing the height index ySbIdx from 0 to numSbY. Processing is performed while changing the width direction index xSbIdx from 0 to numSbX.

[0265] If the top left position of the coding block colCb is (xCb, yCb), then the left The top position (xSb, ySb) is calculated as follows:

[0266] xSb = xCb + xSbIdx * sbWidth ySb = yCb + ySbIdx * sbHeight

[0267] Next, the position where the temporal motion vector tempMv is added to the predicted sub-block colSb is The position of the top left corner of the predicted sub-block colSb is (xColSb, yColSb) (S4154). If the temporal motion vector tempMv is (tempMv[0], tempMv[1]) with 1 / 16 pixel accuracy, The top left position of colSb is as follows:

[0268] xColSb = Clip3( xCtb, xCtb + CtbSizeY + 3, xSb + ( tempMv[0] >> 4 ) ) yColSb = Clip3( yCtb, yCtb + CtbSizeY - 1, ySb + ( tempMv[1] >> 4 ) )

[0269] Here, the top left position of the treeblock is (xCtb, yCtb), and the size of the treeblock is CtbS. As shown in the above formula, the position after adding tempMv is significantly different from before adding tempMv. If this position is outside the screen, it is corrected to a range of about the size of the tree block. If this happens, the image will be corrected within the screen.

[0270] Then, inter prediction information is derived for each reference list (S4156, S4158). Here, for the prediction sub-block colSb, the movement of each reference list in sub-block units is A vector mvLXSbCol and a flag availableFlagLXSbCo indicating whether the prediction sub-block is available or not LX indicates the reference list, and in the derivation of reference list 0, LX becomes L0, and the reference list In the derivation of step 1, LX becomes L1. The derivation of the inter prediction information is performed in steps S4022 and S4023 of FIG. 3, so the explanation is omitted.

[0271] After deriving the inter prediction information (S4156, S4158), the predicted sub-block colSb is valid. It is determined whether availableFlagL0SbCol=0 and availableFlagL1SbCol=0 (S4160). If colSb is invalid, it is determined that it is valid otherwise. If colSb is invalid (S4160:N 0), the motion vector mvLXSbCol is set as the central motion vector ctrMvLX (S4162). Then, a flag predFlagLXSbCol indicating whether LX prediction is used is set to the central motion vector. The flag ctrPredFlagLX is set to the sub-block motion information ctrPredFlagLX (S4162). Finish the derivation of

[0272] Again, refer to FIG. 50. Then, the motion vector mvL0SbCol of L0 and the motion vector mvL0SbCol of L1 are The tor mvL1SbCol is calculated as the sub-block merge mode in the sub-block merge mode derivation unit 304. The candidate is added to the merge candidate list subblockMergeCandList (S4028). This addition is made when the flag availableSbCol=1 indicates the existence of sub-block time merge candidates. This is the only step that is performed by the temporal merge candidate derivation unit 342.

[0273] The above description of the sub-block temporal merge candidate derivation unit 381 is for the encoding process. The same applies to decoding. The operation of the sub-block temporal merge candidate derivation unit 481 is the same as the decoding of the encoding in the above description. and is explained in the same way.

[0274] <Motion compensation prediction processing> The motion compensation prediction unit 306 predicts the current block being predicted during encoding. The motion compensation prediction unit 306 also obtains the inter prediction information. The inter prediction information is acquired from the inter prediction mode determination unit 305. The image data and motion vectors are derived and stored in the decoded image memory 104 as specified by the reference index. The reference picture to be used is moved from the same position as the image signal of the prediction block by the amount of the motion vector. After acquiring the image signal at the position where the image signal is input, a predicted signal is generated.

[0275] In inter prediction, the inter prediction mode is a single reference picture, such as L0 prediction or L1 prediction. In the case of prediction from a reference picture, the prediction signal obtained from one reference picture is used as the motion compensated prediction signal. The inter prediction mode is BI prediction, and the prediction mode is from two reference pictures. In the case of prediction, the weighted average of the predicted signals obtained from two reference pictures is used as the motion vector. The motion compensation prediction signal is supplied to the prediction method determination unit 105. The ratio of the weighted average of the predictions is 1:1, but weighted averages can also be performed using other ratios. For example, if the picture interval between the picture to be predicted and the reference picture is close, The weighting ratio may be increased as the value of the pixel increases. This may be done using a correspondence table of combinations of channel intervals and weighting ratios.

[0276] The motion compensation prediction unit 406 has the same function as the motion compensation prediction unit 306 on the encoding side. The compensation prediction unit 406 outputs the inter prediction information to the normal prediction motion vector mode derivation unit 401, A normal merge mode derivation unit 402, a sub-block predicted motion vector mode derivation unit 403, a sub-block predicted motion vector mode derivation unit 404, a sub-block predicted motion vector mode derivation unit 405, a sub-block predicted motion vector mode derivation unit 406, a sub-block predicted motion vector mode derivation unit 407, a sub-block predicted motion vector mode derivation unit 408, a sub-block predicted motion vector mode derivation unit The block merge mode is obtained from the block merge mode derivation unit 404 via the switch 408. The prediction unit 406 supplies the obtained motion compensation prediction signal to the decoded image signal superimposing unit 207 .

[0277] <About inter prediction mode> The process of making predictions from a single reference picture is defined as uni-prediction, and in the case of uni-prediction, it is called L0 prediction. Or L1 prediction, which is a prediction using one of the two reference pictures registered in the reference lists L0 and L1. Predictions are made using either one of them.

[0278] FIG. 32 shows a uni-predictive picture in which the L0 reference picture (RefL0Pic) is the picture to be processed. 33 shows the case where the time is before CurPic. 1. Similarly, the case where the reference picture to be processed is later than the current picture is shown in FIG. 32 and 33, the reference picture for L0 prediction is changed to the reference picture for L1 prediction (RefL1Pi It is also possible to replace it with c) and perform uni-prediction.

[0279] The process of making predictions from two reference pictures is defined as bi-prediction, and in the case of bi-prediction, L0 prediction is used. The bi-prediction is expressed as BI prediction using both L0 and L1 prediction. The reference picture for L1 prediction is located before the current picture, and the reference picture for L1 prediction is located before the current picture. Figure 35 shows a case where the picture is bi-predictive and has a reference to L0 prediction. This indicates that the reference picture and the reference picture for L1 prediction are located at a time earlier than the picture being processed. FIG. 36 shows a bi-predictive coding example in which a reference picture for L0 prediction and a reference picture for L1 prediction are processed. This shows the case where the picture is located after the picture to be processed.

[0280] In this way, the relationship between the L0 / L1 prediction type and time is that L0 is the past direction and L1 is the future direction. In the case of bi-prediction, the same reference picture can be used. It is also possible to perform L0 prediction and L1 prediction using the same. The decision as to whether to use L0 prediction or bi-prediction is made based on, for example, whether to use L1 prediction or not. The determination is made based on information (for example, a flag) indicating whether or not the message is being sent.

[0281] <About reference indexes> In the embodiment of the present invention, in order to improve the accuracy of the motion compensation prediction, multiple This allows the optimum reference picture to be selected from a number of reference pictures. The reference picture used in the motion compensation prediction is used as a reference index, and The index is coded into the bitstream along with the differential motion vector.

[0282] <Motion compensation processing based on normal predicted motion vector mode> The motion compensation prediction unit 306 is also shown as the inter prediction unit 102 on the encoding side in FIG. In this way, the inter prediction mode determination unit 305 performs normal prediction motion vector mode derivation. When the inter prediction information by the unit 301 is selected, the inter prediction information is - The inter prediction mode of the block currently being processed is acquired from the prediction mode determination unit 305. The estimation mode, reference index, and motion vector are derived to generate a motion compensated prediction signal. The generated motion compensation prediction signal is supplied to the prediction method determination unit 105.

[0283] Similarly, the motion compensation prediction unit 406 also performs the same function as the inter prediction unit 203 on the decoding side in FIG. As shown, during the decoding process, the switch 408 is turned on to the normal predicted motion vector mode derivation unit 40 When connected to 1, inter prediction by the normal prediction motion vector mode derivation unit 401 The information is acquired and the inter prediction mode and reference index of the currently processed block are The generated motion compensation prediction signal is then used to derive a motion vector. The signal is supplied to a decoded image signal superimposing unit 207.

[0284] <Motion compensation processing based on normal merge mode> The motion compensation prediction unit 306 is also shown as the inter prediction unit 102 on the encoding side in FIG. In this way, the inter prediction mode determination unit 305 determines whether the normal merge mode is When inter prediction information by the inter prediction mode is selected, this inter prediction information is used as the inter prediction mode. the inter prediction mode of the block currently being processed, obtained from the prediction mode determination unit 305; A reference index and a motion vector are derived, and a motion compensation prediction signal is generated. The compensated prediction signal is supplied to the prediction method determination unit 105 .

[0285] Similarly, the motion compensation prediction unit 406 also performs the same function as the inter prediction unit 203 on the decoding side in FIG. As shown, during the decoding process, switch 408 is normally connected to merge mode derivation unit 402. If so, the inter prediction information is obtained from the normal merge mode derivation unit 402, and the current processing The inter prediction mode, reference index, and motion vector of the block being processed are The generated motion compensation prediction signal is used as a decoded image signal. The paper is supplied to the folding section 207.

[0286] <Motion compensation processing based on sub-block predicted motion vector mode> The motion compensation prediction unit 306 is also shown as the inter prediction unit 102 on the encoding side in FIG. In this way, the inter prediction mode determination unit 305 determines the sub-block predicted motion vector model. When the inter prediction information is selected by the code derivation unit 303, this inter prediction information The inter prediction mode determination unit 305 obtains the inter prediction mode of the block currently being processed. Derives the center prediction mode, reference index, and motion vector, and generates a motion compensated prediction signal. The generated motion compensation prediction signal is supplied to the prediction method determination unit 105.

[0287] Similarly, the motion compensation prediction unit 406 also performs the same function as the inter prediction unit 203 on the decoding side in FIG. As shown, during the decoding process, a switch 408 controls the sub-block predicted motion vector mode. When the sub-block prediction motion vector mode derivation unit 403 is connected to the sub-block prediction motion vector mode derivation unit 403, Inter prediction information is obtained by the current block being processed. The generated code, reference index, and motion vector are derived to generate a motion compensated prediction signal. The resulting motion compensation prediction signal is supplied to the decoded image signal superimposing unit 207 .

[0288] <Motion compensation processing based on sub-block merge mode> The motion compensation prediction unit 306 is also shown as the inter prediction unit 102 on the encoding side in FIG. In this way, the inter prediction mode determination unit 305 includes a sub-block merge mode derivation unit When the inter prediction information by 304 is selected, this inter prediction information is Inter prediction of the block currently being processed, obtained from the prediction mode determination unit 305 The mode, reference index, and motion vector are derived, and a motion compensation prediction signal is generated. The resulting motion compensation prediction signal is supplied to the prediction method determination unit 105.

[0289] Similarly, the motion compensation prediction unit 406 also performs the same function as the inter prediction unit 203 on the decoding side in FIG. As shown, during the decoding process, a switch 408 controls the sub-block merge mode derivation unit 404. , the inter prediction information by the sub-block merge mode derivation unit 404 is and obtains the inter prediction mode and reference index of the currently processed block. , a motion vector is derived, and a motion compensation prediction signal is generated. The generated motion compensation prediction signal is , and is supplied to the decoded image signal superimposing unit 207.

[0290] <Motion compensation processing based on affine transformation prediction> In normal motion vector prediction mode and normal merge mode, the following flags are used: The following flags are used to indicate whether motion compensation is performed in the encoding process. The following flags are used based on the inter-prediction conditions determined by the inter-prediction mode determination unit 305: The bitstream is then coded into the bitstream. Determine whether to perform motion compensation using an affine model based on the following flags in the program: .

[0291] sps_affine_enabled_flag is the flag for motion compensation using the affine model in inter prediction. If sps_affine_enabled_flag is 0, it indicates whether or not affine is available. The motion compensation is suppressed so that it is not based on the affine model. cu_affine_type_flag specifies the CU (Coded Block) syntax of the coded video sequence. If sps_affine_enabled_flag is 1, the encoded video sequence is not transmitted. In this case, affine model motion compensation can be used.

[0292] sps_affine_type_flag is the 6-parameter affine model for inter prediction. Indicates whether motion compensation is available. If sps_affine_type_flag is 0, 6 parameters are used. It is suppressed so that it is not motion compensation by the affine model. Also, cu_affine_type_flag is not conveyed in the CU syntax of a coded video sequence. If e_flag is 1, the coded video sequence is generated using a 6-parameter affine model. If sps_affine_type_flag is not present, it is assumed to be 0. do.

[0293] When decoding a P or B slice, the current CU If r_affine_flag is 1, generate a motion compensation prediction signal for the CU currently being processed. To do this, affine motion compensation is used. If inter_affine_flag is 0, If this is set, no affine model is used for the current CU. If lag is not present, it is assumed to be 0.

[0294] When decoding a P or B slice, in the currently processed CU, If ffine_type_flag is 1, the motion compensation prediction signal for the CU currently being processed is generated. To achieve this, a six-parameter affine model of motion compensation is used. If e_flag is 0, the motion compensation prediction signal for the CU currently being processed is generated. In this paper, motion compensation using a four-parameter affine model is used.

[0295] In motion compensation using the affine model, reference indices and motion vectors are calculated for each subblock. Since the torque is derived, the reference index that is processed in sub-block units is and motion vectors to generate a motion compensated prediction signal.

[0296] The four-parameter affine model is defined as the horizontal and vertical components of the motion vectors of the two control points. The motion vector of a sub-block is derived from the four vertical component parameters. This is a mode in which motion compensation is performed at the pixel position.

[0297] (Second embodiment) In the second embodiment, the sub-block merge mode derivation units 304 and 404 When deriving affine construction merge candidates, each affine control point (CP0, CP1, CP2, CP3 ) Is it possible to fix the reference image to be compared with the reference image in (a) and perform an affine transformation? For example, the reference image with the reference image index 0 is fixed and each affine Affine transformation is possible with each combination of control points (CP0, CP1, CP2, CP3). Determine whether or not this is the case.

[0298] By comparing the reference image of each affine control point with a fixed reference image, the upper-left affine The upper right affine control point CP1, the lower left affine control point CP2, and the lower right affine control point CP3 are used without using the control point CP0. A three-fin affine control point affine model can be constructed using the fin control point CP3. This increases the possibility of deriving affine construction merge candidates, and This improves prediction accuracy and coding efficiency.

[0299] All of the above-described embodiments may be combined in multiple ways.

[0300] In all the embodiments described above, the bitstream output by the image coding device is a specific data so that it can be decoded according to the encoding method used in the embodiment. The image decoding device corresponding to this image coding device has the following data format. It is possible to decode bitstreams of a particular data format.

[0301] To exchange bitstreams between the image coding device and the image decoding device, a wired or When a wireless network is used, the data is bit-coded into a data format suitable for the transmission mode of the communication channel. In this case, the bitstream output by the image coding device may be converted and transmitted. The stream is converted into coded data in a format suitable for the transmission mode of the communication channel and transmitted over the network. a transmitting device that receives coded data from a network and decodes it into a bit stream; The image encoding device includes a receiving device that transmits the image data to the image decoding device. Memory to buffer the output bitstream and packetize the bitstream a packet processing unit and a transmitter for transmitting packetized encoded data over a network; The receiving device receives packetized encoded data via a network. a receiving unit for receiving the encoded data, a memory for buffering the received encoded data, and a packet and a packet processing unit that processes the data to generate a bitstream and provides the bitstream to the image decoding device.

[0302] In addition, by adding a display unit for displaying the image decoded by the image decoding device to the configuration, In this case, the display unit may be a device that displays the decoded image signal generated by the decoded image signal superimposing unit 207. The decoded image signal stored in the decoded image memory 208 is read out and displayed on the screen.

[0303] In addition, by adding an imaging unit to the configuration and inputting the captured image to the image encoding device, In this case, the imaging unit inputs the captured image signal to the block division unit 101. To exert effort.

[0304] FIG. 37 shows an example of the hardware configuration of the encoding / decoding device according to this embodiment. The image encoding device includes the configuration of an image encoding device and an image decoding device according to an embodiment of the present invention. The encoding / decoding device 9000 includes a CPU 9001, a codec IC 9002, an I / O O interface 9003, memory 9004, optical disk drive 9005, network It has a network interface 9006 and a video interface 9009, and each part is connected via a bus. Connected by 9010.

[0305] The image encoding unit 9007 and the image decoding unit 9008 are typically implemented by a codec IC 9002 and The image encoding process of the image encoding device according to the embodiment of the present invention is implemented as an image encoding The image decoding process is executed by the decoding unit 9007 in the image decoding device according to the embodiment of the present invention. The decoding process is performed by the image decoding unit 9008. The I / O interface 9003 For example, a USB interface is used to connect an external keyboard 9104 and mouse 91 The CPU 9001 receives input via the I / O interface 9003. The encoding / decoding device 90 executes the operation desired by the user based on the user's operation. 00. User operations using the keyboard 9104, mouse 9105, etc. , select whether to perform encoding or decoding functions, set the encoding quality, and There are input / output destinations for programs, input / output destinations for images, etc.

[0306] When the user desires to play back images recorded on the disk recording medium 9100 The optical disc drive 9005 reads the bitstream from the inserted disc recording medium 9100. The read bitstream is then sent to the codec I via bus 9010. The image decoder 9008 of the C9002 receives the input bitstream. The image decoding process is performed on the image data in the image decoding device according to the embodiment of the present invention. The image is sent to an external monitor 9103 via a video interface 9009. The encoding / decoding device 9000 has a network interface 9006. It is possible to connect to an external distribution server 9106 and a mobile terminal 9107 via the network 9101. The user converts the image recorded on the disk recording medium 9100 to the image data recorded on the distribution server 9106. If you want to play back images recorded on the mobile terminal 9107, The interface 9006 receives a bitstream from the input disc recording medium 9100. Instead of reading out the bit stream, the bit stream is acquired from the network 9101. When the user desires to play back an image recorded in memory 9004, 004, an image decoding apparatus according to an embodiment of the present invention The image decoding process is performed in the above.

[0307] The user captures an image using an external camera 9102 and encodes it into memory 9004. When an operation is desired, the video interface 9009 receives an image from the camera 9102. The image data is input and sent to the image encoding unit 9007 of the codec IC 9002 via the bus 9010. The image encoding unit 9007 encodes the image input via the video interface 9009. An image encoding process is performed in the image encoding device according to the embodiment of the present invention, and a bitstream is The bitstream is then sent to memory 9004 via bus 9010. The user changes the memory 9004 and sends the bit stream to the disk recording medium 9100. If you wish to record a program, the optical disc drive 9005 will The bitstream is written to the disc recording medium 9100.

[0308] A hardware configuration that has an image encoding device but does not have an image decoding device, or a hardware configuration that has an image decoding device However, it is also possible to realize a hardware configuration that does not include an image coding device. The hardware configuration is, for example, a codec IC 9002, an image encoding unit 9007, or This is realized by replacing the image decoding unit 9008 with the image decoding unit 9009.

[0309] The above encoding and decoding processes are carried out by hardware-based transmission, storage, and reception devices. It can be realized by using ROM (read only memory) or flash memory. Firmware stored in memory, etc., and software executed by a computer, etc. The firmware program and software program may be executed on a computer. The information may be provided by recording it on a recording medium that can be read by a computer or by wired or wireless network. It can be provided from a server via the network, or it can be data from terrestrial or satellite digital broadcasting. It may also be provided as a broadcast.

[0310] The present invention has been described above based on the embodiments. The embodiments are merely examples, and the respective structures thereof are not intended to be limiting. The fact that various variations are possible in the combination of components and treatment processes, and that such variations It will be understood by those skilled in the art that the embodiments are within the scope of the present invention. [Explanation of symbols]

[0311] 100 image encoding device, 101 block division unit, 102 inter prediction unit, 103 intra prediction unit, 104 decoded image memory, 105 prediction method determination unit, 10 6 residual generation unit, 107 orthogonal transformation and quantization unit, 108 bit string encoding unit, 10 9 inverse quantization and inverse orthogonal transformation unit, 110 decoded image signal superposition unit, 111 coding information storage unit 200 image decoding device; 201 bit string decoding unit; 202 block Division unit, 203 inter prediction unit, 204 intra prediction unit, 205 coding information storage Memory 206 inverse quantization and inverse orthogonal transformation unit, 207 decoded image signal superposition unit, 208 Decoded image memory.< / poc>

Claims

1. 1. A storage method for storing a bitstream generated in accordance with a moving image coding method that performs affine transformation on a coding block basis on a recording medium, the moving image coding method comprising: an affine inheritance merge candidate derivation step of deriving affine inheritance merge candidates that inherit the affine model of blocks spatially adjacent to the encoding target block; an affine construct merge candidate derivation step of deriving affine construct merge candidates from a plurality of pieces of motion information of blocks spatially or temporally adjacent to the encoding target block; an affine-fixed merge candidate derivation step of deriving affine-fixed merge candidates in which motion information of affine control points is fixed; and The affine construction merge candidate derivation step fixes an index of a reference image to be compared to 0 and determines whether or not the image is affine transformable. A storage method characterized by:

2. 1. A transmission method for transmitting a bitstream generated according to a video coding method that performs affine transformation on a coding block basis, the video coding method comprising: an affine inheritance merge candidate derivation step of deriving affine inheritance merge candidates that inherit the affine model of blocks spatially adjacent to the encoding target block; an affine construct merge candidate derivation step of deriving affine construct merge candidates from a plurality of pieces of motion information of blocks spatially or temporally adjacent to the encoding target block; an affine-fixed merge candidate derivation step of deriving affine-fixed merge candidates in which motion information of affine control points is fixed; and The affine construction merge candidate derivation step fixes an index of a reference image to be compared to 0 and determines whether or not the image is affine transformable. A transmission method characterized by:

Images