Image decoding device, image decoding method, and program product

Abstract

An image decoding device (200) according to the present invention is provided with: an intra prediction unit (204) that generates first prediction pixels on the basis of decoded pixels and control information; a motion compensation unit (208) that generates second prediction pixels on the basis of the accumulated decoded pixels and control information; a synthesis unit (205) that synthesizes, on the basis of the control information, any combination including at least one of the first prediction pixels and the second prediction pixels into small regions divided by a plurality of line segments, as third prediction pixels; and an adder (206) that adds any one of the first prediction pixel, the second prediction pixel, and the third prediction pixel to the prediction residual to obtain a decoded pixel.

Description

Technical Field

[0001] The present invention relates to an image decoding device, an image decoding method and a program product. Background Art

[0002] Non-patent documents 1 to 3 disclose a geometric partitioning mode (GPM). GPM divides a rectangular block into two with an inclined dividing line, and performs motion compensation (inter-frame prediction) or intra-frame prediction on each of the two divided small areas.

[0003] Specifically, in GPM, after generating motion compensation (inter prediction) pixels and intra prediction pixels based on the motion vectors or intra prediction modes of the divided small regions, the two prediction pixels are synthesized by weighted averaging according to the distance from the dividing line.

[0004] Prior art literature

[0005] Non-patent literature

[0006] Non-patent document 1: ITU-T H.266 / VVC

[0007] Non-patent document 2: Algorithm description of Enhanced Compression Model 5 (ECM5)

[0008] Non-patent document 3: EE2-2.7: GPM adaptive blending (JVET-Z0059, JVET-Z0137), JVET-AA0058 Summary of the invention

[0009] Problems to be solved by the invention

[0010] The rectangular block segmentation method based on the geometric segmentation mode of Non-Patent Document 1 and Non-Patent Document 2 is defined by a straight line, and 64 segmentation methods are pre-set through the combination of different angles and positions.

[0011] For example, in the case where the decoding target block includes a boundary between the foreground and the background, if the boundary is a straight line, it can be separated into two small areas, and high-efficiency encoding can be achieved through appropriate prediction methods.

[0012] However, there is a problem that there is room for improvement in the coding performance when the boundary is not a straight line. Therefore, the present invention is made in view of the above-mentioned problems, and its object is to provide an image decoding device, an image decoding method and a program product with high coding efficiency.

[0013] Means for solving problems

[0014] The first technical solution of the present invention is an image decoding device, the main purpose of which is to have: a decoding unit, which decodes control information and a quantization value; an inverse quantization unit, which inverse quantizes the quantization value as a transform coefficient; an inverse transformation unit, which inverse transforms the transform coefficient as a prediction residual; an intra-frame prediction unit, which generates a first predicted pixel based on the decoded pixel and the control information; an accumulation unit, which accumulates the decoded pixel; a motion compensation unit, which generates a second predicted pixel based on the accumulated decoded pixel and the control information; a synthesis unit, which synthesizes any combination of the first predicted pixel and at least one of the second predicted pixel into a small area divided by multiple line segments as a third predicted pixel according to the control information; and an adder, which adds any one of the first predicted pixel, the second predicted pixel and the third predicted pixel to the prediction residual to obtain the decoded pixel.

[0015] The second technical solution of the present invention is an image decoding method, which mainly comprises: a process of decoding control information and a quantization value; a process of inversely quantizing the quantization value as a transform coefficient; a process of inversely transforming the transform coefficient as a prediction residual; a process of generating a first predicted pixel based on the decoded pixel and the control information; a process of accumulating the decoded pixel; a process of generating a second predicted pixel based on the accumulated decoded pixels and the control information; a process of synthesizing any combination of at least one of the first predicted pixel and the second predicted pixel into a small area divided by multiple line segments as a third predicted pixel according to the control information; and a process of adding any one of the first predicted pixel, the second predicted pixel and the third predicted pixel to the prediction residual to obtain the decoded pixel.

[0016] The third technical solution of the present invention is a program product, which includes a program for making a computer function as an image decoding device. The main point of the program product is that the image decoding device has: a decoding unit, which decodes control information and a quantization value; an inverse quantization unit, which inversely quantizes the quantization value as a transform coefficient; an inverse transformation unit, which inversely transforms the transform coefficient as a prediction residual; an intra-frame prediction unit, which generates a first predicted pixel based on the decoded pixel and the control information; an accumulation unit, which accumulates the decoded pixel; a motion compensation unit, which generates a second predicted pixel based on the accumulated decoded pixel and the control information; a synthesis unit, which synthesizes any combination of at least one of the first predicted pixel and the second predicted pixel into a small area divided by multiple line segments as a third predicted pixel according to the control information; and an adder, which adds any one of the first predicted pixel, the second predicted pixel and the third predicted pixel to the prediction residual to obtain the decoded pixel.

[0017] Effects of the Invention

[0018] According to the present invention, it is possible to provide an image decoding device, an image decoding method, and a program product with high coding efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a diagram showing an example of functional blocks of an image decoding device 200 according to an embodiment.

[0020] Figure 2 This is a diagram showing an example of using the geometric partitioning mode (GPM) as small area partitioning.

[0021] Figure 3 This is a diagram showing an example of a case where a decoding target block is divided by N=2 line segments.

[0022] Figure 4 This is a diagram showing an example of a case where a decoding target block is divided by N=3 line segments.

[0023] Figure 5 The diagram shows examples of selectable division shapes of a decoding target block.

[0024] Figure 6 The diagram shows examples of selectable division shapes of a decoding target block.

[0025] Figure 7 Is assigned to Figure 2 An example of weight coefficients of three patterns of division boundaries of small area B is shown.

[0026] Figure 8This diagram shows an example of a case where a decoding target block is divided by two dividing lines in the vertical and horizontal directions and synthetic prediction is performed in units of predicted pixels according to the distance from the two dividing lines (dividing line 1 and dividing line 2).

[0027] Fig. 9 This is a diagram for explaining an example of a combined area with respect to the dividing line 1 and the dividing line 2.

[0028] Fig.10 This is a flowchart showing an example of the operation of the sorting method for setting the multi-segment division pattern in units of sequences.

[0029] Fig.11 This is a flowchart showing an example of the operation of the sorting method for setting the multi-segment division pattern in units of blocks.

[0030] Fig.12 This is a diagram showing an example of a case where an angular prediction mode parallel to dividing line 1 is applied.

[0031] Fig.13 This is a diagram for explaining the intra-frame prediction mode derivation method involved in this embodiment.

[0032] Fig.14 This is a diagram for explaining the intra-frame prediction mode derivation method involved in this embodiment.

[0033] Fig.15 This is a diagram for explaining the intra-frame prediction mode derivation method involved in this embodiment.

[0034] Fig.16 This is a diagram for explaining the intra-frame prediction mode derivation method involved in this embodiment.

[0035] Fig.17 This is a diagram showing an example of a method for deriving an intra-frame prediction mode based on adjacent reference blocks relative to a normal intra-frame prediction involved in non-patent literature 1 and non-patent literature 2, and a method for deriving an intra-frame prediction mode based on adjacent reference blocks relative to a geometric partitioning mode involved in the present embodiment to which the above-mentioned derivation method is applied.

[0036] Fig.18 This is a diagram showing an example of the correspondence relationship among cu_div_idx, divDirectionIdx, and divLocationIdx.

[0037] Fig.19 This is a flowchart showing a modified example of the operation of the sorting method for setting the multi-segment division pattern in units of sequences.

[0038] Fig. 20This is a flowchart showing a modified example of the operation of the sorting method for setting a division mode including a multi-segment division mode in units of blocks.

[0039] Explanation of symbols

[0040] 200: image decoding device;

[0041] 201: decoding unit;

[0042] 202: Anti-quantization Department;

[0043] 203: reverse transformation unit;

[0044] 204: intra prediction unit;

[0045] 205: Synthesis Department;

[0046] 206: adder;

[0047] 207: Cumulative Department;

[0048] 208: motion compensation unit;

[0049] 210: code input unit;

[0050] 220: Image output unit. DETAILED DESCRIPTION

[0051] Below, refer to the attached Figure 1 The embodiments of the present invention are described in detail. In addition, the components in the following embodiments can be appropriately replaced with existing components, etc., and various modifications including combinations with other existing components can be performed. Therefore, the content of the technical solution recorded in the claims is not limited based on the description of the following embodiments.

[0052] <First Embodiment>

[0053] Below, refer to Figures 1 to 18 , the image decoding device 200 involved in this embodiment is described. Figure 1 This is a diagram showing an example of functional blocks of the image decoding device 200 according to this embodiment.

[0054] like Figure 1 As shown, the image decoding device 200 has a code input unit 210, a decoding unit 201, an inverse quantization unit 202, an inverse transformation unit 203, an intra-frame prediction unit 204, a synthesis unit 205, an adder 206, an accumulation unit 207, a motion compensation unit 208 and an image output unit 220.

[0055] The code input unit 210 is configured to acquire code information encoded by the image encoding device.

[0056] The decoding unit 201 is configured to decode the control information and the quantized value based on the code information input from the code input unit 210. For example, the decoding unit 201 is configured to output the control information and the quantized value by performing variable length decoding on the code information.

[0057] Here, the quantized value is sent to the inverse quantization unit 202, and the control information is sent to the intra prediction unit 204, the synthesis unit 205, and the motion compensation unit 208. In addition, the control information includes information required for controlling the intra prediction unit 204, the synthesis unit 205, the motion compensation unit 208, etc., and may also include header information such as a sequence parameter set, a picture parameter set, a picture header, or a slice header.

[0058] The inverse quantization unit 202 is configured to inversely quantize the quantized value sent from the decoding unit 201 as a decoded transform coefficient. The transform coefficient is sent to the inverse transform unit 203.

[0059] The inverse transform unit 203 is configured to inversely transform the transform coefficients sent from the inverse quantization unit 202 as a decoded prediction residual. The prediction residual is sent to the adder 206.

[0060] The intra prediction unit 204 is configured to generate a first predicted pixel based on the decoded pixel and the control information sent from the decoding unit 201. Here, the decoded pixel is obtained via the adder 206 and accumulated in the accumulation unit 207. In addition, the first predicted pixel is sent to the adder 206.

[0061] The accumulation unit 207 is configured to cumulatively accumulate the decoded pixels sent from the adder 206. The decoded pixels are referred to by the motion compensation unit 208 via the accumulation unit 207.

[0062] The motion compensation unit 208 is configured to generate a second predicted pixel to be added to the prediction residual by the adder 206 based on the decoded pixel obtained by the reference accumulation unit 207 and the control information decoded by the decoding unit 201. The generated second predicted pixel is sent to the adder 206 or the synthesis unit 205.

[0063] The adder 206 is configured to add any one of the first to third predicted pixels generated from the decoded pixels and the like to the prediction residual sent from the inverse transform unit 203 to obtain a decoded pixel. The decoded pixel is sent to the image output unit 220, the accumulation unit 207, and the intra prediction unit 204.

[0064] The synthesis unit 205 is configured to synthesize any combination including at least one of the first predicted pixels and the second predicted pixels into small areas divided by multiple line segments as third predicted pixels, based on the control information decoded by the decoding unit 201 .

[0065] The following describes the synthesis unit 205 which is a characteristic structure of the image decoding device 200 according to this embodiment.

[0066] The function of the synthesis unit 205 is to divide the decoded target block (small area division) into multiple small areas in a way that the prediction residual can be expressed with a small amount of code when the decoded pixels are calculated by the subsequent addition unit 206, and synthesize (synthesized prediction) the corresponding first predicted pixels or second predicted pixels, thereby predicting the pixels of the decoded target block with high precision.

[0067] Figure 2 An example of using the geometric partitioning mode (GPM) for small area partitioning is shown. Figure 2 In the example, the decoding target block is divided into small area A and small area B by a slanted straight line.

[0068] However, in the geometric segmentation mode, since the segmentation is limited to a straight line, it cannot cope with the situation where the boundary between the foreground and the background is complex, and there is a problem that the encoding efficiency cannot be fully improved.

[0069] (Basic concept of multi-segment segmentation mode)

[0070] In order to solve the above-mentioned problem, in the image decoding device 200 according to the present embodiment, the synthesis unit 205 adopts a procedure of dividing the decoding target block by N (N is a natural number greater than 1) line segments.

[0071] For example, Figure 3 An example of dividing a decoding target block by N=2 line segments is shown. Figure 4 An example of a case where a decoding target block is divided by N=3 line segments is shown.

[0072] Hereinafter, the small area division and synthesis prediction of the decoding target block realized by using multiple (N) line segments will be recorded as "multi-line segment division mode".

[0073] The synthesis unit 205 can determine the division type of the multi-line segment division mode by using the control information. The details will be described later.

[0074] By increasing the number N of line segments in the multi-line segmentation pattern, the prediction accuracy can be improved, and by reducing the number N of line segments in the multi-line segmentation pattern, the amount of code for the control information expressing the segmentation shape can be suppressed.

[0075] The synthesizing unit 205 may set the number N of line segments in the multi-line segment division mode to a fixed value. For example, the synthesizing unit 205 may set N=2, N=3, N=4, N=5, etc. as described above.

[0076] The synthesis section 205 may set the fixed value to a common value without depending on the length of the short side, the length of the long side, the size (area), or the aspect ratio of the decoding target block.

[0077] Alternatively, the synthesis unit 205 may set the fixed value to a different value according to the length of the short side, the length of the long side, the size (area), or the aspect ratio of the decoding target block.

[0078] In addition, the synthesizing unit 205 may set the number N of line segments in the multi-line segment segmentation mode to be variable. For example, the synthesizing unit 205 may determine the number N of line segments in proportion to the length of the short side, the length of the long side, and the size (area) of the decoding target block when setting the number N of line segments in the multi-line segment segmentation mode to be variable.

[0079] Furthermore, the synthesis unit 205 may limit at least one of the positional relationship and the angle of the multi-line segments in order to reduce the amount of code for expressing the control information of the segmentation shape.

[0080] For example, the synthesis unit 205 may define the first line segment as the horizontal direction (0 degrees) or the vertical direction (90 degrees) of the decoding target block. Moreover, the synthesis unit 205 may define the n+1th line segment only as the vertical direction (90 degrees) relative to the nth line segment.

[0081] However, since dividing the decoding target block into four equal parts into a field-like division shape can be achieved by the recursive rectangular block division (quadtree / binarytree / ternarytree division) disclosed in non-patent document 1, the synthesis unit 205 is preferably limited to not being able to select a configuration that can be achieved by the existing block division.

[0082] Applying the above definition means, for example, when the number of line segments N in the multi-line segmentation mode is 2, Figure 5 Among the patterns (32 kinds of patterns) shown for dividing in the horizontal direction and the vertical direction by a plurality of dividing points, a dividing shape (multi-line segments) of the decoding target block is selected.

[0083] Alternatively, applying the above limitation means that, when the first line segment can be in an inclined direction (45 degrees) in addition to the horizontal direction (0 degrees) and the vertical direction (90 degrees), Figure 5 The patterns shown and Figure 6 In the pattern of division in the oblique direction shown, the division shape (multi-line segments) of the decoding target block is selected.

[0084] Here, by reducing the above-mentioned restrictions, the effect of improving the prediction accuracy can be obtained, and by increasing the above-mentioned restrictions, the effect of reducing the amount of code for expressing the control information of the segmentation shape can be obtained.

[0085] The synthesizing unit 205 may set the above-mentioned limiting method either fixedly or variably.

[0086] In addition, the synthesis unit 205 may set different limitations according to the size of the decoding target block (the length of the short side, the length of the long side, the size (area), or the aspect ratio, etc.).

[0087] For example, for Figure 5 The division points shown are, when the division points are respectively arranged with a constant number of pixels, the number of division points is proportional to the length of the side of the decoding target block.

[0088] On the contrary, the synthesis unit 205 can set the number of division points fixedly by changing the arrangement of the division points without depending on the size of the decoding target block.

[0089] For example, when the synthesis unit 205 fixedly configures K (K is a natural number) division points along the vertical or horizontal direction of the decoding target block, the division points can be configured as L / (K+1) pixels as the ratio of the length L (L is a natural number) of the side of the decoding target block to the pixel.

[0090] Here, as shown in Non-Patent Document 1, L may be a natural number of 4 or more that is a power of 2, such as 4, 8, 16, 32, 64, or 128. Similarly, K may be a natural number of 4 or more that is a power of 2, such as 4, 8, 16, 32, 64, or 128.

[0091] In addition, regarding the possible value of K, L / (K+1) pixels can be limited to a natural number that is a power of 2 and is greater than or equal to 4, such as 4, 8, 16, 32, 64, or 128.

[0092] Furthermore, when the aspect ratio of the decoding target block is different, the synthesis unit 205 may set different K division points for the vertical and horizontal sides (vertical width and horizontal width) of the decoding target block.

[0093] In addition, the synthesis unit 205 may Figure 5 The center of the decoding target block is contained in the segmentation point.

[0094] Increasing the number of division points can improve the prediction accuracy, while reducing the number of division points can reduce the amount of code that expresses the division shape.

[0095] (Method for Generating Third Prediction Pixel in Multi-line Segment Division Mode)

[0096] Next, a method for generating the third predicted pixel by the synthesis unit 205 in the multi-line segment division mode will be described.

[0097] The synthesis unit 205 is configured to perform weighted averaging (ie, synthesis prediction) of predicted pixels of the small area A and the small area B with respect to the decoding target block divided by multiple line segments, according to the distances from the dividing lines.

[0098] The types of predicted pixels of the small area A and the small area B may be a combination of different inter-frame prediction pixels as in the geometric partitioning mode disclosed in non-patent document 1, or a combination of inter-frame prediction pixels and intra-frame prediction pixels as in the intra-frame prediction in the geometric partitioning mode disclosed in non-patent document 2, or a combination of different intra-frame prediction pixels.

[0099] Here, different inter-frame prediction pixels are generated based on different motion vectors, and different intra-frame prediction pixels are generated by different intra-frame prediction modes.

[0100] For the synthesis prediction of the small area A and the small area B implemented by the synthesis unit 205, a weighted average based on the distance from the dividing line as used for the synthesis prediction of the geometric partitioning mode disclosed in Non-Patent Document 1 or Non-Patent Document 3 can be used.

[0101] The total value of the weight coefficients for the plurality of predicted pixels is pre-set to be 1 for each pixel, and the plurality of predicted pixels are synthesized by weighted averaging using the weight coefficients, and the result is used as the predicted pixel of the synthesizing unit 205 .

[0102] Here, since the pixel with the weight coefficient set to 1 (i.e., the maximum value) adopts the input prediction pixel, and the pixel with the weight coefficient set to 0 (i.e., the minimum value) does not adopt the input prediction pixel, as a concept, it is equivalent to dividing the unit block into multiple small areas and determining which pixel of the multiple input prediction pixels is applied where and at what ratio.

[0103] Figure 7 Indicates that the Figure 2 An example of weight coefficients of three patterns of division boundaries of the small area B shown. Figure 7 In the example of , the distance in pixel units from the position of the segmentation boundary (dividing line) is taken as the horizontal axis, and the vertical axis represents the weight coefficient.

[0104] Specifically, for the distances a and b of the predicted pixel units at the position of the preset segmentation boundary, the following are prepared: a pattern (1) with a range weight coefficient of [0, 1] assigned to [a, b]; similarly, the distances a and b are set to be doubled, respectively, and assigned to a pattern (2) with a range weight coefficient of [0, 1] assigned to [2a, 2b]; similarly, the distances a and b are set to be 1 / 2, respectively, and assigned to a pattern (3) with a range weight coefficient of [0, 1] assigned to [a / 2, b / 2].

[0105] In the case where these weight coefficients are defined as xc, yc uniquely determined by the distance d(xc, yc) from the dividing boundary (dividing line), multiple patterns (variable values) are prepared instead of a limited pattern (fixed value) equal to the width τ of the dividing boundary of the small area disclosed in non-patent document 3, that is, the weight coefficient is a width other than the minimum value or the maximum value.

[0106] Here, xc and yc are coordinates in the decoding target block. That is, the synthesis unit 205 may be configured to set a plurality of weight coefficients according to the distance between pixels of the distance division boundary.

[0107] In addition, the weight coefficient symmetrical to the division boundary may be set to a=b. That is, the synthesizing unit 205 may be configured to set the weight coefficient symmetrical to the division boundary to the above-mentioned weight coefficient. According to the above configuration, b is not required, so the amount of code can be reduced.

[0108] In addition, the weight coefficient asymmetric to the segmentation boundary may be set to a≠b. That is, the synthesis unit 205 may be configured to set the weight coefficient asymmetric to the segmentation boundary to the above-mentioned weight coefficient. According to the above configuration, when there are different blurs on both sides of the boundary, high-precision prediction is possible.

[0109] In addition, the number of a and b is not limited to two, and the weight coefficient can be set by increasing the number of multi-line segments. That is, the synthesis unit 205 can be configured to set the weight coefficient by multi-line segments according to the distance between pixels of the distance segmentation boundary. According to the above structure, when blurring occurs nonlinearly, it is possible to predict with high accuracy.

[0110] In this way, by setting a plurality of weight coefficients according to the distance between pixels of the distance division boundary, it is possible to obtain an effect that the block size can be uniformly derived even for various block sizes such as 8×8 or 64×16.

[0111] The synthesizing unit 205 can arbitrarily set the type, shape, and number of the above-mentioned patterns.

[0112] For example, in the above example, 2 times and 1 / 2 times of the distance a and b are described as various patterns, but it can also be 4 times or 1 / 4 times. In addition, in the above formula, the weight coefficient is set to a value of 0 to 8, but it can also be set to other values ​​such as 0 to 16 or 0 to 32. In particular, when the distance between pixels of the distance division boundary is 2 times or 4 times, by increasing the maximum value of the weight coefficient, the weighted average of the pixel unit can be made more accurate.

[0113] The synthesis unit 205 can select settings of multiple weight coefficients (width of the synthesis area, maximum value and minimum value of the weight coefficient) for small area A and small area B from pre-prepared combinations based on the control information sent from the decoding unit 201.

[0114] For example, the synthesis unit 205 can use the control information to select the above-mentioned weight coefficient from a plurality of pre-prepared patterns (five patterns of 1 / 4 width, 1 / 2 width, 1 width, 2 width and 4 width in non-patent document 3) as disclosed in non-patent document 3.

[0115] Alternatively, the synthesis unit 205 can use the control information to reduce the selectable candidates from a plurality of pre-prepared pattern candidates according to the size of the decoded target block, the length of the short side, the length of the long side or the aspect ratio, and select the synthesis width indicated by the control information as disclosed in non-patent document 3.

[0116] Here, in the multi-segment division mode, unlike the synthesis prediction for the geometric division mode disclosed in Non-Patent Documents 1 to 3, there are a plurality of division boundaries (division lines) for dividing the small regions, and therefore synthesis prediction is performed for these plurality of division lines.

[0117] Figure 8 Indicates Figure 5 As shown, an example of a case where two dividing lines in the vertical and horizontal directions constitute a division boundary with respect to the decoding target block, and synthetic prediction is performed in units of predicted pixels according to the distance from the two dividing lines (dividing line 1 and dividing line 2).

[0118] The synthesis unit 205 may apply a common weighted average to the synthesis predictions with respect to the plurality of dividing lines.

[0119] Alternatively, the combining unit 205 may apply a weighted average consisting of the maximum value of the different weight coefficients and the distance from the dividing line.

[0120] Moreover, the synthesis unit 205 can be configured to select a weight coefficient from multiple weight coefficients based on at least one of the length of the short side, the length of the long side, the aspect ratio, the size (number of pixels) or the type of division mode of the decoding target block.

[0121] Alternatively, the synthesis unit 205 may be configured to select a weight coefficient from a plurality of weight coefficients in accordance with the type of the intra prediction mode when intra prediction is applied to the small region A or the small region B.

[0122] Alternatively, the synthesis unit 205 may be configured to select a weight coefficient from a plurality of weight coefficients according to the size (number of pixels) of the small area A or the small area B divided in the decoding target block by a multi-line segment division mode.

[0123] Alternatively, the synthesis unit 205 may be configured to select a weighting factor from a plurality of weighting factors according to the size (number of pixels) of the dividing line.

[0124] Alternatively, the synthesis unit 205 may be configured to select a weight coefficient from a plurality of weight coefficients according to a ratio with a side of the decoding target block in the same direction when the direction of the dividing line is horizontal or vertical with respect to the decoding target block.

[0125] In addition, in the multi-segment segmentation mode, due to the fact that there are multiple segments constituting the segmentation boundary, the area where the weight coefficient in the weighted average is the maximum or minimum value, that is, the area of ​​multiple predicted pixels with different synthesized predictions (synthesized area) sometimes repeats in the vertical direction of each segmentation line.

[0126] Fig. 9 It is used to explain the composite area relative to the dividing line 1 and the dividing line 2 ( Fig. 9 A diagram of an example of the gray area in (a).

[0127] like Fig. 9 As shown in (a), a region where the composite regions of the dividing line 1 and the dividing line 2 overlap (hereinafter, the overlapping composite region) is indicated by a dotted frame.

[0128] In addition, Fig. 9 (b) shows an example of a weight coefficient W_1A applied to the predicted pixels of the small area A with respect to the dividing line 1 and an example of a weight coefficient W_2A applied to the predicted pixels of the small area A with respect to the dividing line 2.

[0129] The weight coefficients W_1B and W_2B for each small area B are obtained by subtracting W_1A and W_2A from the maximum value of the weight coefficients.

[0130] The synthesis unit 205 can be as follows Fig. 9 As shown in calculation example 1 of (c), the third predicted pixels of the synthesis area and the overlapped synthesis area are predictively synthesized using the weight coefficient newly generated by selecting the minimum value of each element W_1A and W_2A of the weight coefficient.

[0131] Alternatively, the synthesis unit 205 may be as follows Fig. 9 As shown in calculation example 2 of (c), the third predicted pixels of the composite area and the overlapped composite area are predictively composited using the weight coefficient newly generated by calculating the product of the elements W_1A and W_2A for calculating the weight coefficient.

[0132] (Whether the multi-segment segmentation mode can be used and how to determine the segmentation type)

[0133] Next, the control information decoded by the decoding unit 201 will be described.

[0134] The code information input to the image decoding apparatus 200 may include a sequence parameter set (SPS) in which control information of a sequence unit is collected. In addition, the code information may include a picture parameter set (PPS) or a picture header (PH) in which control information of a picture unit is collected. Furthermore, the code information may include a slice header (SH) in which control information of a slice unit is collected.

[0135] Reference Fig.10 , an example of the operation of a sorting method for setting a multi-segment segmentation pattern in units of sequences is described.

[0136] like Fig.10 As shown, in step S101, the decoding unit 201 determines whether sps_div_enabled_flag is 1 in the sequence parameter set.

[0137] Here, sps_div_enabled_flag is a syntax for controlling whether or not the split mode is used. When sps_div_enabled_flag is 1, it indicates that the split mode is enabled, and when sps_div_enabled_flag is 0, it indicates that the split mode is disabled.

[0138] When sps_div_enabled_flag is 1, the operation proceeds to step S102, and when sps_div_enabled_flag is 0, the operation ends.

[0139] In step S102 , the decoding unit 201 decodes sps_div_multi_flag.

[0140] Here, sps_div_multi_flag is a syntax for controlling the presence or absence of the multi-segment division mode. When sps_div_multi_flag is 1, it indicates that the multi-segment division mode is valid (N>1), and when sps_div_multi_flag is 0, it indicates that the multi-segment division mode is invalid (N=1).

[0141] When sps_div_multi_flag is 1, the operation proceeds to step S103, and when sps_div_multi_flag is 0, the operation ends.

[0142] In step S103, the decoding unit 201 decodes sps_div_multi_mode. Here, sps_div_multi_mode is a syntax for controlling a multi-line segment division mode.

[0143] By using sps_div_multi_mode, the setting of the multi-segment division mode can be changed in units of sequences according to image characteristics, and thus the effect of maximizing the encoding efficiency can be expected.

[0144] For example, since a sequence composed of CG contains more horizontal and vertical boundaries, it can be set in a manner limited to the type of division composed of right angles. Since a sequence composed of natural images can be set in a manner that relaxes the limitation on the type of division, the coding efficiency can be maximized.

[0145] When the method of selecting the boundary width candidates is set in units of pictures, the decoding unit 201 decodes pps_div_enabled_flag, pps_div_multi_flag, and pps_div_multi_mode in the picture parameter set or the picture header.

[0146] By using pps_div_multi_mode, the setting of the multi-line segment division mode can be changed according to the image characteristics in units of pictures, so the effect of maximizing the encoding efficiency can be expected.

[0147] For example, since a screen composed of CG can be set in a manner limited to a segmentation type composed of right angles, and a screen composed of natural images can be set in a manner that relaxes the limitation on segmentation types, the encoding efficiency can be maximized.

[0148] When the method of sorting the boundary width candidates is set in units of slices, the decoding unit 201 decodes sh_div_enabled_flag, sh_div_multi_flag, and sh_div_multi_mode in the slice header in the same manner.

[0149] By using sh_div_multi_mode, the setting of the multi-segment division mode can be changed in units of slices according to image characteristics, and thus the effect of maximizing the encoding efficiency can be expected.

[0150] For example, since the slice area containing a partial image composed of CG can be set in a manner limited to a segmentation type composed of right angles, and the slice area containing a natural image can be set in a manner that relaxes the limitation on the segmentation type, the coding efficiency can be maximized.

[0151] The increase in the amount of code can be suppressed by setting only in the upper layer, and adaptive control can be performed by giving priority to the setting in the lower layer in addition to the setting in the lower layer.

[0152] Alternatively, when the multi-line segment division pattern is set in advance, the decoding of the multi-line segment division pattern itself can be omitted.

[0153] In the above example, a method of setting the multi-segment division mode in units of sequences, screens, or slices is described, but it is also possible to set the multi-segment division mode directly in units of blocks described later without setting these.

[0154] In this case, the degree of freedom in setting the multi-segment division mode decreases, but the increase in the header information described above can be avoided.

[0155] Reference Fig.19 , describing an example of changing the action of the sorting method for setting a multi-segment segmentation pattern in units of sequences.

[0156] like Fig.19 As shown, Fig.19 and Fig.10 The difference is that it includes step S104.

[0157] In step 104, the decoding unit 201 determines whether the technology related to the reordering of the partitioning mode corresponding to the decoded value of cu_div_idx (control information) for matching a specific partitioning mode according to the template described later in the sequence parameter set (hereinafter referred to as the reordering of the partitioning mode subscript according to the template) is valid by determining whether the sps_div_template_reordering_enabled_flag (control information) controlled in sequence units is 1.

[0158] Here, when sps_div_template_reordering_enabled_flag is 1, it indicates that the subscript reordering according to the division mode of the template is enabled, and when sps_div_template_reordering_enabled_flag is 0, it indicates that the subscript reordering according to the division mode of the template is disabled.

[0159] In addition, the value of sps_div_template_reordering_enabled_flag is estimated by the decoding unit 201 before step S104 or without being decoded.

[0160] When sps_div_template_reordering_enabled_flag is not decoded, the decoding unit 201 estimates the value of sps_div_template_reordering_enabled_flag to be 0.

[0161] When sps_div_template_reordering_enabled_flag is 1 (Yes), the decoding unit 201 proceeds to step S102. When sps_div_template_reordering_enabled_flag is 0 (No), the decoding unit 201 ends the processing.

[0162] The details will be described later, but rearranging the subscripts according to the segmentation mode of the template has the effect of shortening the code length of the subscripts in the segmentation mode. Therefore, the multi-segment division mode is determined to be valid only when the subscript rearrangement according to the segmentation mode of the template is valid, thereby reducing the code amount of cu_div_idx for specifying the type of multi-segment division mode or the type of division mode including the multi-segment division mode in units of target blocks, and as a result, it is expected that the encoding performance will be improved.

[0163] Reference Fig.11 , an example of the operation of the sorting method of setting the multi-segment division pattern in units of blocks is described.

[0164] like Fig.11 As shown, in step S201 , the decoding unit 201 determines whether sps_div_enabled_flag, pps_div_enabled_flag, and sh_div_enabled_flag are all 1.

[0165] When neither is 1, the action ends; when both are 1, the action proceeds to step S202.

[0166] In step S202 , the decoding unit 201 determines whether the decoding target block is in the division mode.

[0167] If yes, the action proceeds to step S203, and if no, the action ends.

[0168] In step S203 , the decoding unit 201 decodes cu_div_idx which is a control signal indicating the division mode.

[0169] cu_div_idx is decoded in such a manner as to specify one of the candidates of the multi-line segment division mode selected in div_multi_mode applied to the lowermost layer of the decoding target block.

[0170] The decoding unit 205 decodes the above-mentioned cu_div_idx, and specifies the multi-line segment division mode based on the decoded value.

[0171] For example, prepare Figure 5 The 32 segmentation pattern candidates of the multi-line segmentation pattern shown in FIG. Fig.18The decoded values ​​of cu_div_idx for the 32 patterns shown.

[0172] Each decoded value of cu_div_idx is used for a specific Figure 5 The internal parameter divDirectionIdx of the division direction (which of the upper left, upper right, lower right or lower left has the division boundary (area divided by the division line)) of the four patterns of the multi-segment division mode shown corresponds to the internal parameter divLocationIdx that determines the division points of the eight patterns.

[0173] When the multi-line segment division mode is additionally applied to the geometric division mode of Non-Patent Document 1, it can be realized by creating a corresponding table of 64 decoded values ​​of patterns in which cu_div_idx is added to merge_gpm_partition_idx of a specific geometric division mode.

[0174] merge_gpm_partition_idx is associated with angleIdx indicating the angles of 20 patterns of the dividing lines of 64 patterns for expressing the geometric partition mode and distanceIdx indicating the distances of the four patterns.

[0175] When the multiline segment division mode is applied in addition to the geometric division mode, if divDirectionIdx and divLocationIdx are configured as internal parameters as new values ​​of angleIdx and distanceIdx, the decoding unit 201 can specify the pattern of the multiline segment division mode in addition to the geometric division mode by decoding merge_gpm_partition_idx.

[0176] Reference Fig. 20 , an example of the operation of a sorting method for setting a division pattern including a multi-segment division pattern in units of blocks is described.

[0177] like Fig. 20 As shown, Fig. 20 and Fig.11 The difference is that it includes step S204 and step S205.

[0178] In step 104, the decoding unit 201 determines whether a predetermined condition is satisfied. If the decoding unit 201 determines that the predetermined condition is satisfied, the process proceeds to step S205, and if the decoding unit 201 determines that the predetermined condition is not satisfied, the process proceeds to step S203.

[0179] Here, the predetermined condition may include a condition that the block size of the target block is equal to or smaller than (or smaller than) a predetermined block size. The predetermined block size may be specified by a power of 2 pixel number such as 8×8 pixels, 16×16 pixels, 32×32 pixels, 64×64 pixels, or 128×128 pixels.

[0180] Due to the nature of using multiple line segments to segment the target block, in large-sized blocks, the segmentation boundaries of the multiple line segments are not easy to coincide with the block boundaries within the block.

[0181] On the other hand, since the multi-segment segmentation boundaries in small-sized blocks are easily consistent with the block boundaries within the block, by setting the block size threshold in a manner that limits large-sized blocks as described above, the multi-segment segmentation mode can be made effective only for small-sized blocks, resulting in improved encoding performance.

[0182] As a modified example, the threshold value determination may be performed based on the short side of the target block instead of the block size of the target block.

[0183] Specifically, the condition of step S204 may include a condition that the short side of the target block is less than (or smaller than) a predetermined number of pixels. The predetermined number of pixels may be specified by a power of 2 such as 8 pixels, 16 pixels, 32 pixels, 64 pixels, or 128×128 pixels.

[0184] This can achieve the same effect as the above-mentioned threshold determination based on the block size of the target block.

[0185] On the contrary, the predetermined condition may include a condition that the block size of the target block is equal to or larger than a predetermined block size. The predetermined block size may be specified by a pixel number that is a power of 2, such as 4×4 pixels, 8×8 pixels, 16×16 pixels, or 32×32 pixels.

[0186] Due to the property of using multiple line segments to divide the target block, as mentioned above, in a small-sized block, the segmentation boundary of the multiple line segments is easy to coincide with the block boundary within the block. However, in an extremely small block size, since the distance between the multiple line segments or between the multiple line segments and the target block boundary is short, it is not easy to produce a difference from the existing segmentation mode or the existing coding block segmentation that uses a single line segment to divide the target block.

[0187] Therefore, with respect to extremely small block sizes, as described above, by performing threshold judgment based on the block size, the multi-segment segmentation mode with respect to the target block of an extremely small size can be invalidated, thereby reducing the amount of code for the specific control information required for the multi-segment segmentation mode, thereby improving the encoding performance.

[0188] As a modified example, the threshold value determination may be performed based on the length of the target block instead of the block size of the target block. Specifically, the condition specified in step S204 may include the condition that the long side of the target block is more than (or greater than) a specified pixel. The specified pixel may be specified by a power of 2 such as 4 pixels, 8 pixels, 16 pixels, or 32 pixels.

[0189] This can achieve the same effect as the above-mentioned threshold determination based on the block size of the target block.

[0190] The decoding unit 201 determines whether the technology related to the reordering of the division pattern corresponding to the decoded value of cu_div_idx (control information) for matching a specific division pattern according to a template described later in the sequence parameter set (hereinafter referred to as the division pattern subscript reordering according to the template) is valid by controlling whether the sps_div_template_reordering_enabled_flag (control information) controlled in sequence units is 1.

[0191] When sps_div_template_reordering_enabled_flag is 1, it indicates that the subscript reordering according to the partition mode of the template is enabled, and when sps_div_template_reordering_enabled_flag is 0, it indicates that the subscript reordering according to the partition mode of the template is disabled.

[0192] When sps_div_template_reordering_enabled_flag is 1 (Yes), the decoding unit 201 proceeds to step S102 .

[0193] On the other hand, when sps_div_template_reordering_enabled_flag is 0 (No), the decoding unit 201 ends the processing.

[0194] The details will be described later, but rearranging the subscripts according to the segmentation mode of the template has the effect of shortening the code length of the subscripts in the segmentation mode. Therefore, only when the subscript rearrangement according to the segmentation mode of the template is effective, the multi-segment segmentation mode is determined to be effective, thereby reducing the code amount of cu_div_idx (control information) for specifying the type of multi-segment segmentation mode or the type of segmentation mode including the multi-segment segmentation mode in units of decoding target blocks, and as a result, it is expected that the encoding performance can be improved.

[0195] (Method for deriving motion information for multi-segment segmentation mode)

[0196] The following describes a method for deriving motion information for the multi-line segmentation mode.

[0197] Motion information regarding small region A or small region B segmented in the multi-segment segmentation mode can apply the same method for deriving motion information as that disclosed in Non-Patent Document 1.

[0198] Specifically, the motion compensation unit 208 creates a motion information candidate list (fusion candidate list) composed of the motion information of neighboring blocks of the decoding target block with respect to small region A and small region B, and uses control information (fusion index) for identifying the motion information within the fusion candidate list conveyed from the image encoding device to derive motion information from the fusion candidate list.

[0199] When both small region A and small region B are in inter-frame prediction, the decoding unit 201 decodes the fusion indexes representing candidates of different motion information in such a way as to derive different motion information respectively.

[0200] In addition, as a method for deriving and registering candidates of motion information in the motion information candidate list, a technique of spatial fusion is disclosed in Non-Patent Document 1. Specifically, the motion information at the positions of A0, A1, B0, B1, and B2 adjacent to the decoding target block shown is registered as candidates (spatial fusion candidates) of the motion information of the decoding target block in the motion information candidate list. Fig.17 The motion compensation unit 208 can limit the spatial fusion candidates that can be registered in the motion information candidate list according to the multi-segment segmentation mode. Specifically, registration can be limited only to the spatial fusions adjacent to each small region divided according to the multi-segment segmentation mode.

[0201] When none of the spatial fusion candidates are adjacent to each small region, the spatial fusion candidates that can be registered can be limited only to the closest spatial fusion candidate, or only to N (N is a natural number, N < M) spatial fusion candidates located in close positions with respect to all M (M is a natural number, 5 in the above example) spatial fusion candidates.

[0202] In addition, when the small region divided by the multi-segment segmentation mode is far from all the spatial fusion candidates, such as at the lower right end of the decoding target block, all the candidates can be included in the registration target with respect to the small region.

[0203] Furthermore, in

[0204] In addition, in Fig.17 As an example of spatial fusion candidates, a total of 5 cases including 1 at the upper left end, 2 at the right end, and 2 at the lower left end of the decoding target block are exemplified, but this technique can also be applied to cases of new spatial fusion candidate positions with respect to Non-Patent Document 1, such as between B2 and B0 and between B2 and A0.

[0205] (Derivation method for intra-frame prediction mode for multi-line segmentation mode)

[0206] The following describes a method for deriving an intra prediction mode for a multi-line segment division mode.

[0207] The synthesis unit 205 may apply a parallel angular prediction mode (Angular prediction mode) to each dividing line for the intra prediction mode for the small area A or the small area B divided by the multi-line segment division mode.

[0208] Alternatively, the synthesis unit 205 may apply a vertical angular prediction mode (Angular prediction mode) to each dividing line for the intra prediction mode for the small area A or the small area B divided by the multi-line segment division mode.

[0209] Fig.12 An example of the case where the angular prediction mode parallel to the dividing line 1 is applied is shown.

[0210] Alternatively, the synthesis unit 205 may derive the intra prediction mode for the small area A or the small area B divided in the multi-segment division mode using the derivation technique based on analysis of adjacent pixels disclosed in Non-Patent Document 2.

[0211] Three derivation techniques are described below.

[0212] [Intra-frame prediction mode derivation method based on adjacent reference pixels 1]

[0213] Below, use Figure 13 to Figure 15 , describing the method for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a usual intra-frame prediction involved in non-patent document 2 and the method 1 for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a geometric partitioning mode involved in the present embodiment to which the said derivation method is applied.

[0214] Fig.13 This is a diagram showing an example of a method for deriving an intra-frame prediction mode based on adjacent reference pixels for normal intra-frame prediction according to non-patent document 2 and a method 1 for deriving an intra-frame prediction mode based on adjacent reference pixels for a geometric partitioning mode according to the present embodiment to which the derivation method is applied. Hereinafter, these derivation methods are collectively referred to as "DIMD (Decoder-side Intra Mode Derivation)".

[0215] In non-patent document 2, for the DIMD, as Fig.13As shown, a horizontal and vertical Sobel filter with a window size of 3×3 pixels is applied to adjacent reference pixels adjacent to the decoded target block, and a histogram of pixel values ​​for all angle prediction modes for normal intra-frame prediction is calculated. Here, the calculation method for the angle and pixel value of adjacent reference pixels used to apply the Sobel filter to establish a corresponding relationship with each angle prediction mode can also adopt the same structure as non-patent document 2 in this embodiment, so the detailed description is omitted.

[0216] In non-patent document 2, according to the block size of the decoding target block, as Fig.13 The adjacent reference pixel region used for calculation of the histogram is controlled as shown in FIG. Specifically, in a 4×4 pixel block, the histogram is calculated using only 3×3 pixel regions above and to the left of the uppermost left pixel of the decoding target block.

[0217] In non-patent document 2, intra-frame prediction pixels are generated using the intra-frame prediction mode and planar mode of the pixel values ​​as the highest point and the second highest point in the calculated histogram, and the generated intra-frame prediction pixels are weighted averaged using specified weights to generate final intra-frame prediction pixels.

[0218] In this embodiment, the synthesis unit 205 may apply the DIMD disclosed in the above-mentioned non-patent document 2 only to the derivation of the intra prediction mode of the multi-line segment division mode. That is, the synthesis / generation process of the intra prediction pixels using the derived multiple intra prediction modes is not performed.

[0219] Therefore, since intra-frame prediction pixels can be generated with one intra-frame prediction mode relative to the intra-frame prediction area of ​​the multi-segment segmentation mode (Intra / Intra (intra-frame / intra-frame) - two intra-frame prediction areas in the case of the multi-segment segmentation mode), it is possible to avoid an increase in the circuit scale required for generating intra-frame prediction pixels of the multi-segment segmentation mode in the image decoding device installed with hardware, and at the same time, by analyzing the histogram of adjacent reference pixels of the decoding target block, it is possible to apply intra-frame prediction of textures such as edges of the segmentation shape suitable for the multi-segment segmentation mode, thereby improving the intra-frame prediction performance, and as a result, it is possible to expect an improvement in the encoding performance.

[0220] In addition, in the present embodiment, similar to Non-Patent Literature 2, the decoding unit 201 may be configured to determine whether to derive the intra prediction mode by decoding or estimating a flag for determining whether DIMD is applicable.

[0221] In addition, the synthesis unit 205 involved in the present embodiment can be configured to register the intra-frame prediction mode derived from the DIMD when the same intra-frame prediction mode is not already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode, and not register the intra-frame prediction mode derived from the DIMD when the same intra-frame prediction mode is already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode.

[0222] According to the above structure, it is possible to avoid duplication of registering the same intra prediction mode in the intra prediction mode candidate list.

[0223] Here, when a new intra-frame prediction mode is registered relative to the intra-frame prediction mode candidate list, the consistency with the existing intra-frame prediction mode is compared, and when the two are consistent, the cropping process is subsequently referred to as "intra-frame prediction mode candidate cropping process".

[0224] In addition, the synthesis unit 205 involved in this embodiment can limit the number of intra prediction modes registered in the intra prediction mode candidate list among the intra prediction modes derived from the DIMD to one. In this case, the synthesis unit 205 derives the angle prediction mode as the highest pixel value (brightness value) from the histogram.

[0225] In addition, in the above-mentioned intra-frame prediction mode candidate cropping process, when cropping the angle prediction mode (hereinafter referred to as the first angle prediction mode) as the highest pixel value (brightness value), it can be compared with the existing intra-frame prediction mode from the side with a higher histogram, and the inconsistent modes can be registered.

[0226] Alternatively, in the above-mentioned intra-frame prediction mode candidate cropping process, when the first angle prediction mode is cropped, the DIMD-based intra-frame prediction mode derivation process may be ended.

[0227] As a modified example, the number of intra prediction modes registered in the intra prediction mode candidate list among the intra prediction modes derived by DIMD may be limited to two. In this case, the synthesis unit 205 derives the first angular prediction mode and the second angular prediction mode which is the next point of the highest pixel value (brightness value) from the histogram.

[0228] In addition, in the case of cropping the first angle prediction mode or the second angle prediction mode, as in the above case, the side with the higher histogram following these can be compared with the existing intra-frame prediction mode, and the inconsistent modes can be registered, or the export processing of the DIMD-based intra-frame prediction mode can be directly terminated.

[0229] Furthermore, the synthesis unit 205 according to this embodiment can limit the adjacent reference pixels used for the above-mentioned histogram calculation of DIMD to a predetermined area according to the segmentation shape of the multi-segment segmentation pattern (ie, the angle of the segmentation lines of the multi-segment segmentation pattern).

[0230] Fig.14 This is a diagram showing a table for limiting the area of ​​a reference template (adjacent reference pixels) according to the segmentation lines of a multi-segment segmentation pattern in the template matching technology for a multi-segment segmentation pattern disclosed in Non-Patent Document 2.

[0231] Specifically, Fig.14 A and L shown represent the upper part and the left part of the decoding target block, respectively.

[0232] In this embodiment, by applying a table of (restricted) adjacent reference pixels specified by the dividing lines of the multi-segment segmentation pattern disclosed in the non-patent document 2 to the calculation of the DIMD histogram, it is possible to avoid using all adjacent reference pixels adjacent to the decoded target block for calculation in the calculation of the DIMD histogram, and at the same time, it is possible to derive angle predictions using adjacent reference pixels that only exist in the direction of the dividing lines of the multi-segment segmentation pattern, thereby reducing the derivation processing load of the DIMD-based intra-frame prediction mode relative to the inter-frame prediction of the multi-segment segmentation pattern.

[0233] Alternatively, the synthesis unit 205 may be as follows Fig.15 As shown, with respect to the small area A or the small area B divided by the multi-line segmentation mode, only the reference pixel area facing the block interface across the left or upper side of the decoding target block is used to derive the intra-frame prediction mode based on DIMD instead of the above-mentioned Fig.14 A table of neighboring reference pixels.

[0234] In addition, the synthesis unit 205 can be in a multi-line segmentation mode, such as Fig.15 As shown in the example of (c), with respect to the small area A or the small area B, in the case where there is no reference pixel area across the left or upper block interface of the decoding target block, as shown in Fig.15 As shown in (c), all reference pixels are used to derive the DIMD-based intra prediction mode.

[0235] [Intra-frame prediction mode derivation method based on adjacent reference pixels 2]

[0236] Below, use Figure 14 to Figure 16 , describing the method for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a usual intra-frame prediction involved in non-patent document 3 and the method 2 for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a geometric partitioning mode involved in the present embodiment to which the said derivation method is applied.

[0237] Fig.16 This is a diagram showing an example of a method for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a normal intra-frame prediction involved in non-patent document 3 and a method 2 for deriving an intra-frame prediction mode based on adjacent reference pixels relative to a geometric partitioning mode involved in the present embodiment to which the derivation method is applied. Hereinafter, these derivation methods are collectively referred to as "TIMD (Template-based Intra Mode Derivation)".

[0238] In non-patent document 3, for TIMD, if Fig.16 As shown, the SATD (Sum of Absolute Transformed Difference: Hadamard transform algorithm) of adjacent reference pixels of a specified line adjacent to the decoding target block (hereinafter referred to as the template), adjacent reference pixels facing the template, and intra-frame prediction pixels facing the template (hereinafter referred to as intra-frame prediction pixels facing the template) generated using a specified intra-frame prediction mode is calculated, and the intra-frame prediction mode with the smallest and second largest SATD among the specified intra-frame prediction modes is exported as the intra-frame prediction mode of TIMD to generate intra-frame prediction pixels.

[0239] Here, the intra prediction mode used in the calculation of the SATD of the above-mentioned TIMD is an intra prediction mode included in the intra prediction mode candidate list for normal intra prediction.

[0240] In the TIMD of non-patent document 3, when the vertical prediction mode, the horizontal prediction mode, and the DC prediction mode are not included in the intra prediction mode candidate list for normal intra prediction, the SATD is calculated to derive the intra prediction mode while including them.

[0241] In this embodiment, the TIMD disclosed in the non-patent document 3 may be applied to derive the intra prediction mode by the synthesis unit 205. That is, the synthesis unit 205 does not perform synthesis / generation processing of intra prediction pixels using the derived plurality of intra prediction modes.

[0242] According to the structure, since intra-frame prediction pixels can be generated with one intra-frame prediction mode relative to the intra-frame prediction area of ​​the multi-segment segmentation mode (two intra-frame prediction areas in the case of the Intra / Intra-multi-segment segmentation mode), it is possible to avoid increasing the circuit scale required for generating intra-frame prediction pixels of the multi-segment segmentation mode in the image decoding device installed with hardware, and at the same time, by analyzing the histogram of adjacent reference pixels of the decoding target block, it is possible to apply intra-frame prediction of textures such as edges of the segmentation shape suitable for the multi-segment segmentation mode, thereby improving the intra-frame prediction performance, and as a result, it is possible to expect improvement in encoding performance.

[0243] In addition, in the present embodiment, similarly to Non-Patent Literature 2, the decoding unit 201 may be configured to determine whether to derive the intra prediction mode by decoding or estimating a flag for determining whether TIMD is applicable.

[0244] In addition, the synthesis unit 205 involved in the present embodiment can be configured to register the intra-frame prediction mode derived from TIMD when the same intra-frame prediction mode is not already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode, and not register the intra-frame prediction mode derived from TIMD when the same intra-frame prediction mode is already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode.

[0245] According to the above structure, it is possible to avoid duplication of registering the same intra prediction mode in the intra prediction mode candidate list.

[0246] Here, when a new intra-frame prediction mode is registered relative to the intra-frame prediction mode candidate list, the consistency with the existing intra-frame prediction mode is compared, and when the two are consistent, the cropping process is subsequently referred to as "intra-frame prediction mode candidate cropping process".

[0247] In addition, the synthesis unit 205 according to the present embodiment may limit the number of intra prediction modes registered in the intra prediction mode candidate list among the intra prediction modes derived from TIMD to one. In this case, the synthesis unit 205 derives the intra prediction mode (angular prediction) with the minimum SATD cost from the calculation of SATD.

[0248] However, in the present embodiment, unlike Non-Patent Document 3, the DC prediction mode may be excluded from the calculation of SATD in the TIMD process.

[0249] This is because the DC prediction of generating intra-frame predicted pixels using all adjacent reference pixels adjacent to the decoding target block cannot appropriately reflect the texture such as the edges of the segmentation shape according to the multi-segment segmentation pattern and may generate intra-frame predicted pixels. Therefore, by excluding the DC prediction from the calculation of SATD processed by TIMD, it is possible to avoid deriving the DC prediction mode in TIMD.

[0250] In addition, in the above-mentioned intra-frame prediction mode candidate cropping process, when the angle prediction mode with the smallest SATD cost (hereinafter referred to as the first angle prediction mode) is cropped, the existing intra-frame prediction mode can be compared with the one with the lower SATD cost in turn, and the inconsistent modes can be registered. Alternatively, when the first angle prediction mode is cropped, the TIMD-based intra-frame prediction mode derivation process can be terminated.

[0251] As a modified example, the number of intra prediction modes registered in the intra prediction mode candidate list among the intra prediction modes derived from TIMD may be limited to two. In this case, the synthesis unit 205 derives the first angular prediction mode and the second angular prediction mode which is the next point of the minimum SATD cost from the SATD cost.

[0252] In addition, in the case of cropping the first angle prediction mode or the second angle prediction mode, as in the above case, the one with the lower SATD cost following these can be compared with the existing intra-frame prediction mode, and the inconsistent modes can be registered, or the export processing of the TIMD-based intra-frame prediction mode can be directly terminated.

[0253] Furthermore, the synthesis unit 205 according to this embodiment can limit the adjacent reference pixels used for calculating the above-mentioned TIMD histogram to a predetermined area according to the segmentation shape of the multi-segment segmentation pattern (ie, the angles of the segmentation lines of the multi-segment segmentation pattern).

[0254] In this embodiment, by Fig.14 The table of adjacent reference pixel areas specified by the dividing lines of the multi-segment segmentation pattern disclosed in the non-patent document 2 shown is applied to the calculation of SATD in TIMD processing, which can avoid using all adjacent reference pixels adjacent to the decoded target block for calculation in the calculation of SATD in TIMD processing, and at the same time can use adjacent reference pixels that only exist in the direction of the dividing lines of the multi-segment segmentation pattern to derive angle prediction, thereby reducing the load of the derivation processing of the TIMD-based intra-frame prediction mode relative to the inter-frame prediction of the multi-segment segmentation pattern.

[0255] Furthermore, the synthesis unit 205 according to this embodiment may be configured not to process the SATD calculation of the intra prediction mode when the same intra prediction mode as the intra prediction mode used for calculating the SATD in the TIMD process is already registered in the intra prediction mode candidate list.

[0256] According to the above configuration, since it is possible to avoid repeatedly registering the same intra prediction mode through TIMD processing for an intra prediction mode already registered in the intra prediction mode, it is possible to reduce the load of TIMD processing for inter prediction in the multi-segment division mode.

[0257] Alternatively, the synthesis unit 205 may be as follows Fig.15 As shown, with respect to the small area A or the small area B divided by the multi-line segmentation mode, only the reference pixel area across the left or upper block interface of the decoding target block is used to derive the intra-frame prediction mode based on TIMD instead of the above-mentioned Fig.14 The adjacent reference pixel table shown.

[0258] In addition, the synthesis unit 205 can be in a multi-line segmentation mode, such as Fig.15 As shown in (c), with respect to the small area A or the small area B, in the case where there is no reference pixel area across the left or upper block interface of the decoding target block, as shown in Fig.15 As shown in (c), all reference pixels are used to derive the TIMD-based intra prediction mode.

[0259] [Intra-frame prediction mode derivation method based on adjacent reference blocks]

[0260] Below, use Fig.14 , 17 , describing the method for deriving an intra-frame prediction mode based on adjacent reference blocks relative to a normal intra-frame prediction involved in non-patent literature 1 and non-patent literature 2, and the method for deriving an intra-frame prediction mode based on adjacent reference blocks relative to a geometric partitioning mode involved in the present embodiment to which the above-mentioned derivation method is applied.

[0261] Fig.17 This is a diagram showing an example of a method for deriving an intra-frame prediction mode based on an adjacent reference block relative to a normal intra-frame prediction involved in non-patent literature 1 and non-patent literature 2 and a method for deriving an intra-frame prediction mode based on an adjacent reference block relative to a geometric partitioning mode involved in this embodiment to which the above-mentioned deriving method is applied. Hereinafter, these deriving methods are collectively referred to as "BIMD (Block-based Intra Mode Derivation)".

[0262] In non-patent document 3, for BIMD, if Fig.17 As shown, the intra-frame prediction mode of the adjacent reference block at a specified position adjacent to the decoding target block is derived as the intra-frame prediction mode of the BIMD to generate intra-frame prediction pixels. In addition, when the adjacent reference block is an intra-frame prediction block, the intra-frame prediction mode of the adjacent reference block derived here directly refers to the intra-frame prediction mode of the adjacent reference block, but when the adjacent reference block is an inter-frame prediction block or an inter-frame prediction block and a multi-line segmentation mode application block to which intra-frame prediction is applied, the intra-frame prediction mode stored in units of 4×4 sub-block pixels described later is referenced.

[0263] Here, in non-patent document 1 and non-patent document 2, as Fig.17 As shown, the adjacent reference blocks referenced in the above BIMD are set to the left (A0), lower left (A1), top (B0), upper right (B1) and upper left (B2) of the decoding target block.

[0264] In this embodiment, the BIMD disclosed in Non-Patent Documents 1 and 2 can be applied to derive the intra prediction mode by the synthesis unit 205. That is, the synthesis unit 205 does not perform synthesis / generation processing of intra prediction pixels using the derived plurality of intra prediction modes.

[0265] According to the structure, since it is possible to select an intra-frame prediction mode from an intra-frame prediction mode candidate list of intra-frame prediction modes that may include an adjacent reference block of a decoding target block to generate intra-frame prediction pixels with respect to the intra-frame prediction area of ​​the multi-segment segmentation mode (two intra-frame prediction areas in the case of the Intra / Intra-multi-segment segmentation mode), it is possible to apply intra-frame prediction of textures such as edges of a segmentation shape suitable for the multi-segment segmentation mode, thereby improving the intra-frame prediction performance and, as a result, expecting an improvement in the encoding performance.

[0266] In addition, the synthesis unit 205 involved in the present embodiment can be configured to register the intra-frame prediction mode derived from BIMD when the same intra-frame prediction mode is not already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode, and not register the intra-frame prediction mode derived from BIMD when the same intra-frame prediction mode is already included in the intra-frame prediction mode candidate list for the multi-segment segmentation mode.

[0267] According to the above structure, it is possible to avoid duplication of registering the same intra prediction mode in the intra prediction mode candidate list.

[0268] Here, when a new intra-frame prediction mode is registered with respect to the intra-frame prediction mode candidate list, the consistency with the existing intra-frame prediction mode is compared, and when the two are consistent, the cropping process is subsequently referred to as "intra-frame prediction mode candidate cropping process".

[0269] Furthermore, unlike Non-Patent Document 1 and Non-Patent Document 2, the synthesis unit 205 according to the present embodiment may be configured not to register the intra prediction mode candidate list when the intra prediction mode derived from the BIMD is a DC prediction mode.

[0270] This is because the DC prediction for generating intra-frame prediction pixels using all adjacent reference pixels adjacent to the decoding target block cannot appropriately reflect the texture such as the edges of the segmentation shape according to the multi-segment segmentation pattern and may generate intra-frame prediction pixels. Therefore, by excluding DC prediction from the intra-frame prediction mode of BIMD, it is possible to avoid using the DC prediction mode in the generation of intra-frame prediction pixels.

[0271] In addition, in the derivation of the intra-frame prediction mode by the synthesis unit 205 and the BIMD according to the present embodiment, a reference may be constituted similarly to Non-Patent Documents 1 and 2. Fig.17The order of the maximum five adjacent reference blocks is shown. In addition, since the reference order is disclosed in Non-Patent Document 1 and Non-Patent Document 2, the detailed description is omitted in this embodiment.

[0272] As a modified example, in this embodiment, by Fig.14 The table of (restricted) adjacent reference pixel areas specified by the dividing lines of the multi-segment segmentation pattern disclosed in the non-patent document 2 shown is applied to the reference of the adjacent reference blocks of BIMD, which can avoid referring to the intra-frame prediction modes of all adjacent reference blocks adjacent to the decoding target block in the derivation of the intra-frame prediction mode of BIMD, and at the same time, can derive the intra-frame prediction mode with reference to the intra-frame prediction mode (angular prediction) of the adjacent reference blocks that only exist in the direction of the dividing lines of the multi-segment segmentation pattern, thereby reducing the load of the derivation processing of the BIMD-based intra-frame prediction mode relative to the inter-frame prediction of the multi-segment segmentation pattern.

[0273] [Intra-prediction mode derivation method using control information]

[0274] Regarding the above-mentioned different types of intra-frame prediction modes, the synthesis unit 205 may uniquely apply the above-mentioned intra-frame prediction mode, or may select an actually applied intra-frame prediction mode from a plurality of different intra-frame prediction mode candidates included in the intra-frame prediction mode candidate list according to the control information.

[0275] The following is an example of the structure of a plurality of different intra prediction modes included in the intra prediction mode candidate list. Here, they are described as an angular prediction mode parallel to the dividing line (Parallel) and an angular prediction mode perpendicular to the dividing line (Perpendicular).

[0276] Structure example 1. (or, Perpendicular)

[0277] Structure example 2. (or, Perpendicular)

[0278] Structure example 3. (or, Perpendicular)

[0279] Structure example 4. (or, Perpendicular)

[0280] Structure Example 5. (or, Perpendicular)

[0281] Structure Example 6. (or, Perpendicular)

[0282] Structure Example 7. (or, Perpendicular)

[0283] First, structural examples 1 to 3 are methods of combining Parallel (or Perpendicular) with DIMD, TIMD, and BIMD, respectively.

[0284] The intra-frame prediction mode derived by Parallel (or, Perpendicular) can more easily and directly derive an intra-frame prediction mode that reflects textures such as edges based on dividing lines than the intra-frame prediction modes derived by DIMD, TIMD and BIMD. However, since the possibility of deriving a prediction mode with high precision is lower than DIMD, TIMD and BIMD based on the analysis of adjacent reference pixels, it is arranged in the list after these intra-frame prediction mode candidates.

[0285] Next, configuration examples 4 and 5 are configuration examples in which DIMD is arranged before TIMD or BIMD.

[0286] The reason why DIMD is arranged before TIMD is that the intra-frame prediction mode derivation process of DIMD is lighter than the intra-frame prediction mode derivation of TIMD which includes relatively heavy calculation processes such as calculation of SATD.

[0287] On the other hand, the reason for configuring DIMD before BIMD is that the derivation process of the intra-frame prediction mode of DIMD including the calculation of the histogram is lighter than the derivation of the intra-frame prediction mode of BIMD. However, it is possible that the intra-frame prediction mode derived by DIMD can derive an intra-frame prediction mode that reflects the texture such as the edge of the dividing line based on GPM better through the calculation of the histogram than the intra-frame prediction mode derived by BIMD. Therefore, it can be considered that the effect of improving the intra-frame prediction performance is likely to be higher.

[0288] In the configuration example 6, TIMD is arranged before BIMD. The reason for the arrangement is the same as the reason for arranging DIMD before BIMD described above.

[0289] Structural example 7 is a structural example of all combinations of GIMD, DIMD, TIMD, and BIMD. Based on the above reasons, it can be expected that when the intra-frame prediction mode is derived in this order, an intra-frame prediction mode with high prediction performance can be derived more efficiently.

[0290] (Starting restrictions of each intra prediction mode derivation method)

[0291] The synthesis unit 205 involved in this embodiment starts each derivation process of the intra-frame prediction mode relative to the above-mentioned geometric block partitioning mode at the starting time point of each derivation process when the number of intra-frame prediction mode candidates contained in the intra-frame prediction mode candidate list has not reached the maximum value of the intra-frame prediction mode candidate list size, and does not start each derivation process when the number of candidates reaches the maximum value of the intra-frame prediction mode candidate list size.

[0292] According to the above configuration, it is possible to avoid the execution of unnecessary intra prediction mode derivation processing, and it is possible to expect a reduction in the overall processing load of the synthesis unit 205 .

[0293] (Registration method of intra prediction mode candidate list after intra prediction mode derivation is completed)

[0294] The synthesis unit 205 involved in this embodiment can be configured to not register the specified intra-frame prediction mode when the number of intra-frame prediction mode candidates contained in the intra-frame prediction mode candidate list has not reached the maximum value of the intra-frame prediction mode candidate list size at the completion time point of the derivation process of the intra-frame prediction mode relative to the above-mentioned geometric block partitioning mode, and when the intra-frame prediction mode candidate list already includes the same prediction mode.

[0295] For example, in Structural Examples 1 to 7, when the intra prediction mode derived from DIMD, TIMD, and BIMD is the same as the subsequent Parallel (or Perpendicular), the list size does not reach the maximum value.

[0296] In the above case, the synthesis unit 205 may register the unregistered Perpendicular (or Parallel) mode. Alternatively, the synthesis unit 205 may register the planar mode. Alternatively, the synthesis unit 205 may register the DC mode. Alternatively, the synthesis unit 205 may register the adjacent intra-frame prediction mode of the intra-frame prediction mode initially registered in the intra-frame prediction mode candidate list.

[0297] (Method for storing prediction information for multi-line segmentation and multi-line segmentation mode)

[0298] The intra prediction unit 204 stores the intra prediction mode applied to the small region A or the small region B divided in the multi-line segment division mode in units of sub-blocks of a predetermined size that divide the decoding target block.

[0299] The specified size may be, for example, the minimum size of a coding block, a prediction block, or a transform block. Alternatively, the specified size may be a fixed size such as 2×2 pixels or 4×4 pixels.

[0300] In this way, by storing the intra prediction mode applied to the small area A or the small area B in units of sub-blocks instead of units of decoding target blocks, the intra prediction mode in the synthesis area can be accurately stored.

[0301] The intra prediction unit 204 may store both intra prediction modes of the small area A or the small area B for the sub-block located in the synthesis area. Alternatively, the intra prediction unit 204 may store only the corresponding intra prediction mode belonging to either the small area A or the small area B across the dividing line from the center coordinate of the sub-block.

[0302] The motion compensation unit 208 stores motion information (reference image list, reference image index, motion vector) applied to the small area A or small area B divided in the multi-line segment division mode in units of sub-blocks of a predetermined size that divide the decoding target block.

[0303] The specified size may be, for example, the minimum size of a coding block, a prediction block, or a transform block. Alternatively, the specified size may be a fixed size such as 2×2 pixels or 4×4 pixels.

[0304] In this way, by storing the motion information respectively applied to the small area A or the small area B in units of sub-blocks instead of units of decoding target blocks, the motion information in the combined area can be accurately stored.

[0305] The motion compensation unit 208 may store both the motion information of the small area A or the small area B for the sub-block located in the synthesis area. Alternatively, the motion compensation unit 208 may store only the corresponding motion information belonging to either the small area A or the small area B across the dividing line from the center coordinate of the sub-block.

[0306] Alternatively, when the reference image lists of small area A and small area B are different (i.e., one party observes different frames with reference to the future direction from the frame, and the other party observes different frames with reference to the past direction from the frame), the motion compensation unit 208 can generate and save a new motion vector by weighted averaging the respective motion vectors based on the distances between the frame and the reference frame, as in the double prediction disclosed in non-patent document 1.

[0307] Alternatively, when the reference image lists of small area A and small area B are the same (i.e., one side observes different frames referring to the future direction (or the past direction) from the frame, and the other side also observes different frames referring to the future direction (or the past direction) from the frame), the motion compensation unit 208 may only save the motion vector of small area B. Alternatively, in this case, the motion compensation unit 208 may only save the small area A.

[0308] (How to change the order of subscript codes in multi-line segmentation mode)

[0309] The synthesis unit 205 can establish a corresponding multi-line segment division pattern by rearranging the template matching disclosed in Non-Patent Document 2 with the decoded value of the control information cu_div_idx for specifying the multi-line segment division pattern.

[0310] Specifically, when a multi-segment segmentation pattern is formed by inter-frame prediction, the synthesis unit 205 compares the errors (for example, SAD: Sum of Absolute Difference) of the adjacent pixels (templates) of the decoded target block and the reference block for all multi-segment segmentation patterns. When calculating the SAD, the synthesis unit 205 performs weighted averaging in a manner that extends the segmentation line relative to the adjacent pixels.

[0311] When a multi-segment segmentation pattern is formed by intra-frame prediction, the synthesis unit 205 applies the intra-frame prediction mode to the reference pixels that are one row earlier than the adjacent reference pixels of the decoding target block to generate adjacent pixels, and compares the SAD of the generated adjacent pixels with the adjacent pixels of the decoding target block.

[0312] The synthesis unit 205 utilizes the SAD comparison to rearrange the SAD in ascending order and establish a corresponding multi-segment segmentation pattern corresponding to the decoded value of cu_div_idx, thereby being able to utilize a multi-segment segmentation pattern with high prediction accuracy with a smaller decoded value (code length), thereby resulting in improved coding efficiency.

[0313] Furthermore, the synthesis unit 205 can not only rearrange the multi-segment division pattern corresponding to the decoded value of cu_div_idx in ascending order of SAD, but also exclude the multi-segment division pattern from the candidates for selectable multi-segment division pattern when the SAD reaches a specified number in ascending order.

[0314] For example, as the predetermined number, half the number of candidates of selectable multi-segment division modes or half the total count of the multi-segment division mode and the geometric division mode can be set.

[0315] Thus, since the number of multi-segment division patterns corresponding to cu_div_idx and the number of division pattern candidates are reduced, it can be expected that the code length of cu_div_idx can be further shortened, and as a result, the effect of improving coding efficiency can be obtained.

[0316] According to the image decoding device 200 according to the present embodiment, since decoding is performed by dividing each unit block into small regions composed of multiple line segments, it is possible to improve encoding efficiency.

[0317] (Change Example)

[0318] Furthermore, in the above-described embodiment, an example is given of a case where the area is divided into two small areas by a multi-line segment. However, the present invention is not limited to the above case, and can also be applied to a case where the area is divided into three or more small areas by a multi-line segment.

[0319] In addition, in the above-mentioned embodiment, an example is given of a case where all small areas are divided in a manner that includes the edge of the decoding target block, but the present invention is not limited to the above case, and may also be applied to a case where at least one small area is divided in a manner that does not include the edge of the decoding target block (i.e., at least one small area is divided in a manner that is not connected to the periphery of the decoding target block).

[0320] The above-described image decoding apparatus 200 can be realized as a program that causes a computer to execute each function (each process).

[0321] Industrial Applicability

[0322] In addition, according to this embodiment, for example, since comprehensive service quality can be improved in moving image communications, it can contribute to Goal 9 of the Sustainable Development Goals (SDGs) led by the United Nations, "Improve the infrastructure of the information environment, promote sustainable industrialization and achieve expansion of innovation."

Claims

1. An image decoding device, characterized in that: The image decoding device comprises: A decoding unit, which decodes the control information and the quantized value; a dequantization unit, which dequantizes the quantized value as a transform coefficient; an inverse transform unit, which inversely transforms the transform coefficients as prediction residuals; an intra-frame prediction unit, which generates a first predicted pixel according to the decoded pixel and the control information; an accumulation unit for accumulating the decoded pixels; a motion compensation unit, which generates a second predicted pixel according to the accumulated decoded pixel and the control information; a synthesis unit that synthesizes, according to the control information, any combination of at least one of the first predicted pixel and the second predicted pixel into small areas divided by multiple line segments as third predicted pixels; as well as An adder is configured to add any one of the first predicted pixel, the second predicted pixel, and the third predicted pixel to the prediction residual to obtain the decoded pixel.

2. The image decoding device according to claim 1, characterized in that The synthesis unit sets the number of the line segments to a fixed value.

3. The image decoding device according to claim 1, characterized in that The synthesis unit sets the number of the line segments to be variable.

4. The image decoding device according to claim 1, wherein: The synthesis section defines a positional relationship between the plurality of line segments.

5. The image decoding device according to claim 1, characterized in that: The synthesis section defines angles of the plurality of line segments.

6. The image decoding device according to claim 1, characterized in that: The synthesis portion defines only a first line segment in the horizontal direction or the vertical direction.

7. The image decoding device according to claim 1, characterized in that: The synthesis section defines only the (n+1)th line segment in a perpendicular direction with respect to the nth line segment.

8. The image decoding device according to claim 1, characterized in that: The synthesis unit is limited in such a way that it cannot select a configuration that can be realized by the existing block division.

9. The image decoding device according to claim 1, characterized in that: The synthesis section selects the multi-line segments from a pattern divided in the horizontal direction and the vertical direction at a plurality of division points.

10. The image decoding device according to claim 1, characterized in that: The synthesis unit fixedly sets a method for defining the multi-line segments.

11. The image decoding device according to claim 1, characterized in that: The synthesis unit can variably set a method for defining the multi-line segment.

12. The image decoding device according to claim 1, characterized in that: The synthesis unit sets different methods of limiting the multi-line segments according to the size of the decoding target block.

13. The image decoding device according to claim 1, wherein: The synthesis unit arranges the division points of the small area at constant pixels.

14. The image decoding device according to claim 1, characterized in that: The synthesis unit changes the arrangement of the division points of the small region to fixedly set the number of division points without depending on the size of the decoding target block.

15. The image decoding device according to claim 1, characterized in that: The synthesis section applies a common weighted average to the multi-line segment when weighted averaging respective predicted pixels with respect to a first small area and a second small area of ​​a decoding target block partitioned by the multi-line segment according to the distance from the multi-line segment.

16. The image decoding device according to claim 1, characterized in that: The synthesis section applies a different weighted average to each of the multiline segments when weighted averaging the respective predicted pixels with respect to the first small area and the second small area of ​​the decoding target block divided by the multiline segments according to the distance from the multiline segments.

17. The image decoding device according to claim 15 or 16, characterized in that: The synthesis unit selects a weight coefficient from a plurality of weight coefficients according to at least one of a length of a short side, a length of a long side, a size, an aspect ratio, or a type of a division mode of the decoding target block.

18. The image decoding device according to claim 15 or 16, characterized in that: The synthesis unit selects a weight coefficient from a plurality of weight coefficients according to the type of the intra prediction mode.

19. The image decoding device according to claim 15 or 16, characterized in that: The synthesis unit selects a weight coefficient from a plurality of weight coefficients according to a size of the first small area A or the second small area.

20. The image decoding device according to claim 15 or 16, characterized in that: The synthesis unit selects a weight coefficient from a plurality of weight coefficients according to the number of pixels of the multi-line segment.

21. The image decoding device according to claim 15 or 16, characterized in that: When the direction of the multi-line segment is horizontal or vertical with respect to the decoding target block, the synthesis unit selects a weight coefficient from a plurality of weight coefficients according to a ratio of a side of the decoding target block in the same direction as the direction of the multi-line segment.

22. The image decoding device according to claim 15 or 16, characterized in that: The synthesizing unit newly generates a weight coefficient by using the minimum value of each element of the weight coefficients of each of the multi-line segments.

23. The image decoding device according to claim 15 or 16, characterized in that: The synthesizing unit calculates the weight coefficient of each of the multi-line segments using the product of each element.

24. The image decoding device according to claim 1, characterized in that: The synthesis section derives the intra prediction mode of DIMD using only an area of ​​reference pixels facing across a block interface on the left or above the decoding target block with respect to the first small area or the second small area.

25. The image decoding device according to claim 1, characterized in that: The synthesis unit derives the intra prediction mode of DIMD using all reference pixels when there is no area of ​​reference pixels facing the left or upper block interface of the decoding target block relative to the first small area or the second small area.

26. The image decoding device according to claim 1, characterized in that: The synthesis section derives the intra prediction mode of the TIMD using only an area of ​​reference pixels facing across a block interface on the left or above the decoding target block with respect to the first small area or the second small area.

27. The image decoding device according to claim 1, characterized in that: The synthesis unit derives the intra prediction mode of TIMD using all reference pixels when there is no region of reference pixels facing the first small region or the second small region across the left or upper block interface of the decoding target block.

28. The image decoding device according to claim 1, characterized in that: The synthesis unit applies a table of regions of adjacent reference pixels specified by the division lines of the multi-line segment division pattern to a reference of the adjacent reference block of the BIMD.

29. The image decoding device according to claim 1, characterized in that: The synthesis unit selects an intra prediction mode to be actually applied from a plurality of different intra prediction mode candidates included in the intra prediction mode candidate list.

30. An image decoding method, characterized in that: The image decoding method has: A process of decoding control information and quantized values; The step of dequantizing the quantized value as a transform coefficient; The step of inversely transforming the transform coefficients as prediction residuals; A step of generating a first predicted pixel according to the decoded pixel and the control information; A step of accumulating the decoded pixels; A step of generating a second predicted pixel according to the accumulated decoded pixel and the control information; a step of combining, according to the control information, any combination including at least one of the first predicted pixel and the second predicted pixel into small areas divided by multiple line segments as third predicted pixels; as well as A step of adding the prediction residual to any one of the first predicted pixel, the second predicted pixel, and the third predicted pixel to obtain the decoded pixel.

31. A program product comprising a program for causing a computer to function as an image decoding device, wherein: The image decoding device comprises: A decoding unit, which decodes the control information and the quantized value; a dequantization unit, which dequantizes the quantized value as a transform coefficient; an inverse transform unit, which inversely transforms the transform coefficients as prediction residuals; an intra-frame prediction unit, which generates a first predicted pixel according to the decoded pixel and the control information; an accumulation unit for accumulating the decoded pixels; a motion compensation unit, which generates a second predicted pixel according to the accumulated decoded pixel and the control information; a synthesis unit that synthesizes, according to the control information, any combination of at least one of the first predicted pixel and the second predicted pixel into small areas divided by multiple line segments as third predicted pixels; as well as An adder is configured to add any one of the first predicted pixel, the second predicted pixel, and the third predicted pixel to the prediction residual to obtain the decoded pixel.

Images