IMAGE PROCESSING DEVICE AND IMAGE PROCESSING METHOD

TECHNICAL FIELD

The present technology relates to an image processing device and an image processing method, and more particularly, to an image processing device and an image processing method capable of simplifying processing, for example.

BACKGROUND ART

Joint standardization organization of ITU-T and ISO/IEC, Joint Video Experts Team (JVET) aims to further improve encoding efficiency compared to H.265/HEVC, and is working on standardization of versatile video coding (VVC), which is the next-generation image encoding method.

In the standardization operation of VVC, it has been proposed to perform matrix-based intra-prediction (MIP), which is intra prediction using a matrix operation on a prediction block (for example, see Non-Patent Documents 1 and 2).

In MIP, parameters of a matrix (weight matrix) and a bias (bias vector) obtained by parameter learning are defined, and an operation using the matrix and the bias (parameters) is performed.

CITATION LIST
Non-Patent Document

Non-Patent Document 1: JVET-N0217-v3: CE3: Affine linear weighted intra prediction (CE3-4.1, CE3-4.2) (version 7—date Jan 17, 2019)

Non-Patent Document 2: JVET-00084-v1: 8-bit implementation and simplification of MIP (version 1—date Jun. 18, 2019)

SUMMARY OF THE INVENTION
Problems to be Solved by the Invention

In the MIP operation (operation performed by the MIP), a shift is performed by a predetermined shift amount The shift amount is changed according to MipSizeld indicating a matrix size of the matrix and a mode number of the MIP in order to improve bit accuracy.

As described above, in the MIP operation, since the shift amount is changed according to the MipSizeld and the MIP mode number, the shift amount needs to be set for each combination of the MipSizeld and the mode number of the MIP and the processing is complicated.

The present technology has been made in view of such a situation, and makes it possible to simplify processing.

Solutions to Problems

An image processing device of a first aspect of the present technology includes an intra prediction unit configured to perform matrix intra prediction using a shift amount set to a fixed value when performing the matrix intra prediction, which is intra prediction, on a current prediction block to be encoded, using a matrix operation to generate a predicted image of the current prediction block, and an encoding unit configured to encode the current prediction block using the predicted image generated by the intra prediction unit.

An image processing method of a first aspect of the present technology includes an intra prediction process of performing matrix intra prediction using a shift amount set to a fixed value when performing the matrix intra prediction, which is intra prediction, on a current prediction block to be encoded, using a matrix operation to generate a predicted image of the current prediction block, and an encoding process of encoding the current prediction block using the predicted image generated in the intra prediction process.

In the image processing device and the image processing method according to the first aspect of the present technology, when the matrix intra prediction which is the intra prediction using the matrix operation is performed on the current prediction block to be encoded, the matrix intra prediction is performed by using the shift amount set to the fixed value to generate the predicted image of the current prediction block. Then, the current prediction block is encoded using the predicted image.

An image processing device of a second aspect of the present technology includes an intra prediction unit configured to perform matrix intra prediction using a shift amount set to a fixed value when performing the matrix intra prediction, which is intra prediction, on a current prediction block to be decoded, using a matrix operation to generate a predicted image of the current prediction block, and a decoding unit configured to decode the current prediction block using the predicted image generated by the intra prediction unit.

An image processing method of a second aspect of the present technology includes an intra prediction process of performing matrix intra prediction using a shift amount set to a fixed value when performing the matrix intra prediction, which is intra prediction, on a current prediction block to be decoded, using a matrix operation to generate a predicted image of the current prediction block, and a decoding process of decoding the current prediction block using the predicted image generated in the intra prediction process.

In the image processing device and the image processing method according to the second aspect of the present technology, when the matrix intra prediction which is the intra prediction using the matrix operation is performed on the current prediction block to be decoded, the matrix intra prediction is performed by using the shift amount set to the fixed value to generate the predicted image of the current prediction block. Then, the current prediction block is decoded using the predicted image.

Note that the image processing device may be an independent device or an internal block constituting one device.

In addition, the image processing device can be realized by causing a computer to execute a program. The program can be provided by being recorded on a recording medium or by being transmitted through a transmission medium

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1 is a diagram for describing a first generation method of a predicted image of MIP.

FIG. 2 is a diagram for describing a second generation method of a predicted image of MIP.

FIG. 3 is a diagram illustrating an example of a weight matrix mWeight [i] [j] of (M, m)=(1, 3).

FIG. 4 is a diagram illustrating an example of a weight matrix mWeight [i] [j] of (M, m)=(1, 8).

FIG. 5 is a diagram illustrating an example of a bias vector vBias [j] with MipSizeld=1.

FIG. 6 is a diagram for describing a third generation method of a predicted image of MIP.

FIG. 7 is a diagram for describing a fourth generation method of a predicted image of MIP.

FIG. 8 is a diagram illustrating a shift amount sW described in Reference A.

FIG. 9 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 0) described in Reference A.

FIG. 10 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 1) described in Reference A.

FIG. 11 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 2) described in Reference A.

FIG. 12 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 3) described in Reference A.

FIG. 13 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 4) described in Reference A.

FIG. 14 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 5) described in Reference A.

FIG. 15 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 6) described in Reference A.

FIG. 16 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 7) described in Reference A.

FIG. 17 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 8) described in Reference A.

FIG. 18 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 9) described in Reference A.

FIG. 19 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 10) described in Reference A.

FIG. 20 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 11) described in Reference A.

FIG. 21 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 12) described in Reference A.

FIG. 22 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 13) described in Reference A.

FIG. 23 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 14) described in Reference A.

FIG. 24 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 15) described in Reference A.

FIG. 25 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 16) described in Reference A.

FIG. 26 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(0, 17) described in Reference A.

FIG. 27 is a diagram illustrating a standard variable fO of MipSizeld=0 described in Reference A.

FIG. 28 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 0) described in Reference A.

FIG. 29 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 1) described in Reference A.

FIG. 30 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 2) described in Reference A.

FIG. 31 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 3) described in Reference A.

FIG. 32 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 4) described in Reference A.

FIG. 33 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 5) described in Reference A.

FIG. 34 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 6) described in Reference A.

FIG. 35 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 7) described in Reference A.

FIG. 36 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 8) described in Reference A.

FIG. 37 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(1, 9) described in Reference A.

FIG. 38 is a diagram illustrating a standard variable fO of MipSizeld=1 described in Reference A.

FIG. 39 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 0) described in Reference A.

FIG. 40 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 1) described in Reference A.

FIG. 41 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 2) described in Reference A.

FIG. 42 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 3) described in Reference A.

FIG. 43 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 4) described in Reference A.

FIG. 44 is a diagram illustrating a standard weight matrix mWeight [i] [j] of (M, m)=(2, 5) described in Reference A.

FIG. 45 is a diagram illustrating a standard variable fO of MipSizeld=2 described in Reference A.

FIG. 46 is a diagram illustrating the shift amount sW used in the third generation method.

FIG. 47 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 0) used in the third generation method.

FIG. 48 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 1) used in the third generation method.

FIG. 49 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 2) used in the third generation method.

FIG. 50 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 3) used in the third generation method.

FIG. 51 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 4) used in the third generation method.

FIG. 52 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 5) used in the third generation method.

FIG. 53 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 6) used in the third generation method.

FIG. 54 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 7) used in the third generation method.

FIG. 55 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 8) used in the third generation method.

FIG. 56 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 9) used in the third generation method.

FIG. 57 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 10) used in the third generation method.

FIG. 58 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 11) used in the third generation method.

FIG. 59 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 12) used in the third generation method.

FIG. 60 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 13) used in the third generation method.

FIG. 61 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 14) used in the third generation method.

FIG. 62 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 15) used in the third generation method.

FIG. 63 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 16) used in the third generation method.

FIG. 64 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 17) used in the third generation method.

FIG. 65 is a diagram illustrating a fixed variable fO of MipSizeld =0 used in the third generation method.

FIG. 66 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 0) used in the third generation method.

FIG. 67 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 1) used in the third generation method.

FIG. 68 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 2) used in the third generation method.

FIG. 69 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) used in the third generation method.

FIG. 70 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 4) used in the third generation method.

FIG. 71 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 5) used in the third generation method.

FIG. 72 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 6) used in the third generation method.

FIG. 73 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 7) used in the third generation method.

FIG. 74 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 8) used in the third generation method.

FIG. 75 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 9) used in the third generation method.

FIG. 76 is a diagram illustrating a fixed variable fO of MipSizeld=1 used in the third generation method.

FIG. 77 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 0) used in the third generation method.

FIG. 78 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 1) used in the third generation method.

FIG. 79 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 2) used in the third generation method.

FIG. 80 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 3) used in the third generation method.

FIG. 81 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 4) used in the third generation method.

FIG. 82 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 5) used in the third generation method.

FIG. 83 is a diagram illustrating a fixed variable fO of MipSizeld=2 used in the third generation method.

FIG. 84 is a diagram illustrating a shift amount sW used in a fourth generation method.

FIG. 85 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 0) used in the fourth generation method.

FIG. 86 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 1) used in the fourth generation method.

FIG. 87 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 2) used in the fourth generation method.

FIG. 88 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 3) used in the fourth generation method.

FIG. 89 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 4) used in the fourth generation method.

FIG. 90 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 5) used in the fourth generation method.

FIG. 91 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 6) used in the fourth generation method.

FIG. 92 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 7) used in the fourth generation method.

FIG. 93 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 8) used in the fourth generation method.

FIG. 94 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 9) used in the fourth generation method.

FIG. 95 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 10) used in the fourth generation method.

FIG. 96 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 11) used in the fourth generation method.

FIG. 97 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 12) used in the fourth generation method.

FIG. 98 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 13) used in the fourth generation method.

FIG. 99 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 14) used in the fourth generation method.

FIG. 100 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 15) used in the fourth generation method.

FIG. 101 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 16) used in the fourth generation method.

FIG. 102 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(0, 17) used in the fourth generation method.

FIG. 103 is a diagram illustrating a fixed variable fO of MipSizeld=0 used in the fourth generation method.

FIG. 104 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 0) used in the fourth generation method.

FIG. 105 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 1) used in the fourth generation method.

FIG. 106 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 2) used in the fourth generation method.

FIG. 107 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) used in the fourth generation method.

FIG. 108 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 4) used in the fourth generation method.

FIG. 109 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 5) used in the fourth generation method.

FIG. 110 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 6) used in the fourth generation method.

FIG. 111 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 7) used in the fourth generation method.

FIG. 112 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 8) used in the fourth generation method.

FIG. 113 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(1, 9) used in the fourth generation method.

FIG. 114 is a diagram illustrating a fixed variable fO of MipSizeld=1 used in the fourth generation method.

FIG. 115 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 0) used in the fourth generation method.

FIG. 116 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 1) used in the fourth generation method.

FIG. 117 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 2) used in the fourth generation method.

FIG. 118 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 3) used in the fourth generation method.

FIG. 119 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 4) used in the fourth generation method.

FIG. 120 is a diagram illustrating a fixed weight matrix mWeight [i] [j] of (M, m)=(2, 5) used in the fourth generation method.

FIG. 121 is a diagram illustrating a fixed variable fO of MipSizeld=2 used in the fourth generation method.

FIG. 122 is a block diagram illustrating a configuration example of an embodiment of an image processing system to which the present technology is applied.

FIG. 123 is a block diagram illustrating a configuration example of an encoder 11.

FIG. 124 is a flowchart illustrating an example of encoding processing of the encoder 11.

FIG. 125 is a block diagram illustrating a detailed configuration example of a decoder 51.

FIG. 126 is a flowchart illustrating an example of decoding processing of the decoder 51.

FIG. 127 is a block diagram illustrating a configuration example of an embodiment of a computer to which the present technology is applied.

MODE FOR CARRYING OUT THE INVENTION

The scope disclosed herein is not limited to the contents of embodiments, and the contents of the following references REF 1 to REF 7, which are known at the time of filing, are also incorporated herein by reference. In other words, the contents described in the following references REF 1 to REF 7 are also the basis for judging the support requirements. For example, even when a quad-tree block structure, a quad tree plus binary tree (QTBT) block structure, and a multi-type tree (MTT) block structure are not directly defined in the detailed description of invention, they are within the scope of the present disclosure and shall meet the support requirements of the claims. In addition, for example, even when technical terms such as parsing, syntax, and semantics are not directly defined in the detailed description of the invention, similarly, they are within the scope of the present disclosure and shall meet the support requirements of the claims.

REF 1: Recommendation ITU-T H.264 (04/2017) “Advanced video coding for generic audiovisual services”, April 2017

REF 2: Recommendation ITU-T H.265 (02/2018) “High efficiency video coding”, February 2018

REF 3: Benjamin Bross, Jianle Chen, Shan Liu, Versatile Video Coding (Draft 5), JVET-N1001-v7 (version 7—date May 29, 2019)

REF 4: Jianle Chen, Yan Ye, Seung Hwan Kim, Algorithm description for Versatile Video Coding and Test Model 5 (VTM 5), JVET-N1002-v1

REF 5: JVET-N0217-v3: CE3: Affine linear weighted intra prediction (CE3-4.1, CE3-4.2) (version 7—date Jan. 17, 2019)

REF 6: JVET-M0043-v2: CE3: Affine linear weighted intra prediction (test 1.2.1, test 1.2.2) (version 2—date Jan. 9, 2019)

REF 7: JVET-OOO84-v1: 8-bit implementation and simplification of MIP (version 1—date Jan. 18, 2019)

Speaking of pixels, the adjacent includes not only a case in which one pixel (one line) is adjacent to a current pixel of interest, but also a case in which a plurality of pixels (a plurality of lines) are adjacent. Therefore, the adjacent pixel includes not only a pixel at a position of one pixel directly adjacent to a current pixel, but also pixels at positions of a plurality of pixels continuously adjacent to the current pixel. Further, the adjacent block includes not only a block in a range of one block directly adjacent to the current block of interest, but also a block in a range of a plurality of blocks continuously adjacent to the current block. Further, the adjacent block can also include a block located in the vicinity of the current block, as necessary.

The prediction block means a block (prediction unit (PU)) that is a processing unit upon performing intra prediction, and also includes sub-blocks in the prediction block. When the prediction block, an orthogonal transform block (transform unit (TU)), which is a processing unit for performing orthogonal transform, and an encoding block (coding unit (CU)), which is a processing unit for encoding, are unified into the same block, the prediction block, the orthogonal transform block, and the encoding block mean the same block.

The intra prediction mode comprehensively means variables (parameters) referenced when deriving the intra prediction mode, such as a mode number when performing the intra prediction, an index of the mode number, a block size of the prediction block, and a size of a sub-block which is the processing unit in the prediction block.

A matrix intra prediction mode (MIP mode) comprehensively refers to variables (parameters) referenced when deriving the matrix intra prediction mode, such as a mode number of the MIP, an index of the mode number, a type of matrix used when performing MIP operation, a type of matrix size of the matrix used when performing the MIP operation, and the like.

Changing means changing based on the contents (values, arithmetic expressions, variables, or the like) described in references REF 3 or REF 7. The changing can include changing based on the contents described in the document “JVET-N1001-v7_proposal_text_MIP_8Bit.docx” (hereinafter, also referred to as Reference A) that is published in other documents that are known at the time of filing, for example, a web site (http://phenix.int-evry.fr/jvet/) of “JVET DOCUMENT MANAGEMENT SYSTEM”.

For example, in reference REF 3, 8 and 9 are mixed as the shift amount sW used in the MIP operation, but the changing includes changing 9 of the mixed shift amount sW to 8, changing the MIP operation (matrix operation) including a weight matrix mWeight, changing the MIP operation including a bias vector vBias, or the like.

Further, for example, in Reference A (Reference REF 7), 6, 7, and 8 are mixed as the shift amount sW, but the changing includes changing 7 of the mixed shift amount sW to 6, changing from 8 to 7, changing the MIP operation including the weight matrix mWeight, changing the MIP operation including a variable oW corresponding to the bias vector vBias, or the like. The changing of the MIP operation includes the weight matrix mWeight used for the MIP operation, the bias vector vBias, a variable fO used to obtain the variable oW, or the like.

In the present technology, identification data that identifies a plurality of patterns can be set as a bitstream syntax obtained by encoding an image. The bitstream can include the identification data that identifies various patterns.

As the identification data, for example, the information on the shift amount sW, for example, data indicating whether or not the shift amount sW is the fixed value (set to a fixed value) can be adopted.

When the identification data is included in the bitstream, the decoder that decodes the bitstream can perform processing more efficiently by parsing and referencing the identification data.

FIG. 1 is a diagram for describing a first generation method of a predicted image of the MIP.

The first generation method is a method for generating a predicted image of MIP proposed in Reference REF 3 (JVET-N1001-v7).

In the first generation method, according to Equation (1) described as Equation (8-69) in Reference REF 3, predMip [x] [y] (pixel values) of (some) pixels of the predicted image of the current prediction block which is the prediction block to be encoded/decoded) is generated.

predMip[x] [y]=((ΣmWeight[i][y*incH*predC+x*incW]*p[i])+(vBias[y*incH* predC+x*incW]<<sB)+oW)>>sW (1)

predMip [x] [y] represents pixels (pixel values) in which a horizontal position of the predicted image is x and a vertical position thereof is y. The pixel of the predicted image is also referred to as a predicted pixel.

In Equation (1), a summation Σ represents a summation taken while changing variable i to an integer from 0 to 2*boundarySize-1. The boundarySize is set corresponding to the block size of the current prediction block. Specifically, the boundarySize is set according to Table 8-7 of Reference REF 3 according to MipSizeld which is an identifier of the matrix size of the matrix (weight matrix mWeight [i] [j]) used for the operation (matrix operation) with the pixel of the current prediction block in the MIP. The MipSizeld is set to 0, 1, and 2, respectively, when the matrix size is 4×4, 8×8, and 16×16.

mWeight [i] [j] represents components of a j-th row and an i-th column of the matrix used in the MIP, and is referred to as a weight matrix. The weight matrix mWeight [i] [j] is set according to the MipSizeld and the modeld.

The number of values taken by arguments i and j of the weight matrix mWeight [i] [j] is equal to the number of pixels obtained by averaging in reference REF 5 and the number of pixels predMip [x] [y] of the predicted image, respectively.

In addition, a sum-of-product arithmetic mWeight [i] [y*incH * predC+x*incW]*p [i] performed using the weight matrix mWeight [i] [j] is the matrix operation performed on (pixels of) the current prediction block in the MIP. That is, the sum-of-product mWeight [i] [y* incH*predC+x*incW]*p [i] is a matrix operation to obtain a product of a matrix whose component is the weight matrix mWeight [i] [j] and a vector whose component is a pixel p[i] of the current prediction block.

The modeld is set to any value from 0 to 17 according to Equation (8-64) of reference REF 3 according to the intra prediction mode predModeIntra.

incH and incW are set according to Equations (8-67) and (8-68) of Reference REF 3, respectively, according to PredC.

The PredC is set according to Table 8-7 of reference REF 3 according to the MipSizeld.

p[i] represents (the pixel value of) the i-th pixel obtained by averaging using the pixels of the reference image referenced when generating the predicted image of the current prediction block.

vBias [j] is the bias added to the pixels predMip [x] [y] of the predicted image, and is referred to as the bias vector. The bias vector vBias [j] is set according to the sizeId and the modeld. The number of values taken by the argument j of the bias vector vBias [j] is equal to the number of values taken by the argument j of the weight matrix mWeight [i] [j], that is, the number of pixels predMip [x] [y] of the predicted image.

It is stated that the sizeId takes 0, 1, or 2 in Table 8-2 of reference REF 3.

sB is set according to the number of bits bitdpthY represents the pixel value according to Equation (8-66) of the Reference REF 3

oW is set according to sW (Weight shift) according to Equation (8-65) in Reference REF 3.

sW is set according to the MipSizeld and the modeld according to Table 8-8 of reference REF 3.

Note that A<<B and A>>B indicate that A is shifted left and right by a B bit, respectively.

In the first generation method, sW is a shift amount that shifts ((ΣmWeight [i] [y*incH*predC+x* incW]*p [i])+(vBias [y*incH*predC+x*incW]<<sB)+oW).

According to Table 8-8 of reference REF 3, the shift amount sW is set to one of 8 and 9 according to the MipSizeld and the modeld as illustrated in FIG. 1.

Therefore, in the first generation method, it is necessary to reset the shift amount sW for each combination of the MipSizeld and the modeld, which complicates the processing of generating the predicted image of the MIP.

For example, when the first generation method is implemented by hardware, a selector for switching the shift amount sW is required, which increases the circuit scale. Further, when the first generation method is implemented by software, it is necessary to reference the table specified as Table 8-8 in Reference REF 3, and the processing speed is reduced by that amount.

Therefore, in the present technology, the MIP is performed using a shift amount set to a fixed value. That is, for example, the operation of Equation (1) is changed according to the shift amount sW set to the fixed value, and the predicted image of the MIP is generated according to the changed operation.

FIG. 2 is a diagram for describing a second generation method of a predicted image of MIP.

In the second generation method, the shift amount sW of Equation (1) for generating the predicted image of MIP proposed in reference REF 3 (JVET-N1001-v7) is set to a fixed value, for example, 8. Then, in the second generation method, the operation of Equation (1) is changed according to the shift amount sW=8 set to the fixed value of 8, and the predicted image of the MIP is generated according to the changed operation.

Specifically, in the second generation method, the weight matrix mWeight [i] [j] and/or the bias vector vBias [j] is changed according to the shift amount sW=8 set to a fixed value, and the predicted image of the MIP is generated according to the operation including the changed weight matrix mWeight [i] [j] and/or the bias vector vBias [j].

Here, the combination of the MipSizeld and the modeld is represented as (M, m). M represents the MipSizeld and m represents the modeld.

In addition, the predicted pixel predMip [x] [y] obtained before setting the shift amount sW to the fixed value, in FIG. 2, the predicted pixel predMip [x] [y] obtained when the predicted pixel predMip [x] [y] (of the predicted image of the MIP) obtained according to reference REF 3, that is, the predicted pixel predMip [x] [y] obtained when the shift amount sW of (M, m)=(1, 3) and (M, m)=(1, 8) is 8 is also referred to as a standard predicted pixel predMip[x] [y].

In Table 8-8 of Reference REF 3, when (M, m)=(1, 3) and (M, m)=(1, 8), the shift amount sW is 9, and in other combinations of the MipSizeld and the modeld, the shift amount sW is 8. Therefore, in Table 8-8 of reference REF 3, the shift amount sW of (M, m)=(1, 3) and (M, m)=(1, 8) is changed from 9 to 8, and therefore, the shift amount sW can be fixed to 8 for each combination of the MipSizeld and the modeld.

In the second generation method, the shift amount sW of (M, m)=(1, 3) and (M, m)=(1, 8) is changed from 9 to 8, and for each combination of MipSizeld and the modeld, the shift amount sW is fixed to 8. Then, the weight matrix mWeight [i] [j] and the bias vector vBias [j] of (M, m)=(1, 3) and (M, m)=(1, 8) are changed according to the shift amount sW=8 so that the predicted pixel (hereinafter, also referred to as a fixed predicted pixel) obtained after setting the shift amount sW to the fixed value), here, the predicted pixel predMip[x] [y] obtained according to Equation (1) when the shift amount sW of (M, m)=(1, 3) , and (M, m)=(1, 8) is changed to 8 becomes a value approximate the standard predicted pixel predMip[x] [y].

In the second generation method, the predicted image of the MIP is generated according to the operation of Equation (1) including the weight matrix mWeight [i] [j] changed according to the shift amount sW=8 set to the fixed value as described above and the bias vector vBias [j].

Therefore, in the second generation method, the shift amount sW is fixed regardless of the modeld and the combination of the MipSizeld and the modeld, so the processing of the MIP can be simplified. As a result, it is not necessary to specify the table specified as Table 8-8 in Reference REF 3 in the standard, and the standard can be simplified.

In addition, for example, when the second generation method is implemented by hardware, the selector for switching the shift amount sW is not required, which can suppress the increases in the circuit scale. Furthermore, when the second generation method is implemented by software, it is not necessary to reference the table specified as Table 8-8 in Reference REF 3, and it is possible to suppress the decrease in the processing speed compared to the case of referencing the table.

FIG. 3 is a diagram illustrating an example of a weight matrix mWeight [i] [j] of (M, m)=(1, 3).

Here, when the weight matrix mWeight [i] [j] before setting the shift amount sW to the fixed value, for example, the shift amount sW of (M, m)=(1, 3) specified in Reference REF 3 is 9, the weight matrix mWeight [i] [j] of (M, m)=(1, 3) is referred to as the standard weight matrix mWeight [i] [j] of (M, m)=(1, 3). Also, when the weight matrix mWeight [i] [j] after setting the shift amount sW to the fixed value, for example, the shift amount sW of (M, m)=(1, 3) is changed to 8, the weight matrix mWeight [i] [j] in which the standard weight matrix mWeight[i] [j] of (M, m)=(1, 3) used in Equation (1) is referred to as the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3).

A of FIG. 3 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 3), and B of FIG. 3 is a diagram illustrating an example of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3).

In FIG. 3, at the i+1th value from the left, the j+1th value from the top represent the weight matrix mWeight [i] [j].

In FIG. 3, the fixed weight matrix mWeight [i] [j] (B of FIG. 3) of (M, m)=(1, 3) is changed to almost ½ times the standard weight matrix mWeight [i] [j] (A of FIG. 3) of (M, m)=(1, 3). This is because the shift amount sW performed to the right in the equation (8-69) of Reference REF 3 has been changed (fixed) from 9 to 8.

Note that the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) is not limited to the value illustrated in FIG. 3.

For the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3), the shift amount sW set to the fixed value is used to perform the operation, and can be changed as appropriate within the range in which the technical effect is appropriately exhibited. Further, since the range in which the technical effect is exhibited changes depending on the approximate level to be set, it can be appropriately changed as long as it falls within those ranges. For example, it can be changed within the range of ±1 and can be changed within the range of ±3. Furthermore, it is possible not only to change all values uniformly, but also to change only some values. It is also possible to set the range of values to be changed individually with respect to the existing values.

That is, for example, when the operation of Equation (1) was performed using the shift amount sW set to the fixed value, the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) can be appropriately changed within the range in which the technical effects such as ensuring a predetermined prediction accuracy or higher can be obtained for the predicted image.

In addition, the range (degree) at which the technical effect is exhibited is changed depending on how much the degree to which the fixed prediction pixel predMip [x] [y] approximates the standard prediction pixel predMip [x] [y] (hereinafter, also referred to as the approximation level) is set. The fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) can be changed as appropriate within the range in which the set approximation level can be maintained.

For example, the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) can be changed within the range of ±1 based on the value illustrated in B of FIG. 3, or can be changed within the range of ±3.

The change of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) can be made on all of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3), but can be made only on a part of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3).

In addition, as the range of values in which the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) is changed, it is also possible to adopt a uniform range for all of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) and adopt a separate range for each fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3).

For example, the range of values in which the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 3) is changed can be set individually for each value of the corresponding standard weight matrix mWeight [i] [j].

The same also applies to the fixed weight matrix mWeight [i] [j] other than (M, m)=(1, 3), the bias vector vBias [j] changed according to the shift amount sW set to the fixed value, and the variable fO described later, which is changed according to the shift amount sW set to the fixed value.

FIG. 4 is a diagram illustrating an example of a weight matrix mWeight [i] [j] of (M, m)=(1, 8).

A of FIG. 4 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 8), and B of FIG. 4 is a diagram illustrating an example of the fixed weight matrix mWeight [i] [j] of (M, m)=(1, 8).

In FIG. 4, similarly to FIG. 3, at the i +1th value from the left, the j+1th value from the top are the weight matrix mWeight [i] [j].

In FIG. 4, similar to FIG. 3, the fixed weight matrix mWeight [i] [j] (B of FIG. 4) of (M, m)=(1, 8) is changed to almost ½ times the standard weight matrix mWeight [i] [j] (A of FIG. 4) of (M, m)=(1, 8).

FIG. 5 is a diagram illustrating an example of the bias vector vBias [j] of MipSizeld=1.

Here, the bias vector vBias [j] before setting the shift amount sW to the fixed value, for example, the bias vector vBias[j] of MipSizeld=1 when the shift amount sW of MipSizeld=1 specified in Reference REF 3 is 9 is also referred to as the standard bias vector vBias [j] of MipSizeld=1.

In addition, the bias vector vBias [j] after setting the shift amount sW to the fixed value, for example, the bias vector vBias[j] in which the standard bias vector vBias [j] of MipSizeld=1 used in Equation (1) when the shift amount sW of MipSizeld=1 is changed to 8 is referred to as the fixed bias vector vBias[j] of MipSizeld=1.

A of FIG. 5 is a diagram illustrating the standard bias vector vBias [j] with MipSizeld=1, and B of FIG. 5 is a diagram illustrating an example of the fixed bias vector vBias [j] of MipSizeld=1.

In FIG. 5, at the j+1th value from the left, the k+1th value from the top is the bias vector vBias [j] of (M, m)=(1, k).

In FIG. 5, of the fixed bias vector vBias [j] of (M, m)=(1, 0) to (1, 9), the fixed bias vector vBias[j] of (M, m)=(1, 3) and (1, 8) is changed to almost ½ times the standard bias vector vBias [j] of (M, m)=(1, 3) and (1, 8) so as to be surrounded by a dotted line, and the other fixed bias vector vBias [j] is the same as the standard bias vector vBias [j].

FIG. 6 is a diagram for describing a third generation method of a predicted image of MIP.

In the third generation method, the shift amount sW of the operation for generating the predicted image of the MIP proposed in Reference A (JVET-N1001-v7_proposal_text_MIP_8Bit.docx) (and Reference REF 7) is set to the fixed value, for example, 6. Then, in the third generation method, the operation to generate the predicted image of MIP is changed according to the shift amount sW=6 set to the fixed value of 6, and the predicted image of the MIP is generated according to the changed operation.

Here, in the generation of the predicted image proposed in Reference A, (some) pixels predMip [x] [y] of the predicted image of the current prediction block is generated according to Equation (2) described as Equation (8-69) in Reference A.

predMip[x] [y]=(((ΣmWeight[i ][y*incH*predC+x*incW]*p[i])+oW)>>sW)+dcVal (2)

In Equation (2), a summation Σ represents a summation taken while changing variable i to an integer from 0 to inSize-1. The inSize is set to 2* boundarySize-1 or 2*boundarySize.

The variable oW corresponds to the bias vector vBias [j] in Equation (1). The variable oW is calculated according to Equation (3) described as Equation (8-65) in reference A.

oW=(1<<(sW−1))−fO*(Σp[i]) (3)

Similar to Equation (2), in Equation (3), a summation Σ represents a summation taken while changing variable i to an integer from 0 to inSize-1.

The variable fO is set according to the MipSizeld and the modeld. Since the variable oW corresponding to the bias vector vBias [j] is calculated according to Equation (3) using the variable fO, similar to the variable oW, the variable fO can also correspond to the bias vector vBias [j].

In the generation of the predicted image proposed in Reference A, sW is the shift amount that shifts ((ΣmWeight [i] [y*incH*predC+x*incW]*p[i])+oW).

According to Table 8-8 of reference A, the shift amount sW is set to 6, 7, or 8 according to the MipSizeld and the modeld as illustrated in FIG. 6.

Therefore, in the generation of the predicted image proposed in Reference A, there is a similar problem to the first generation method in that it is necessary to reset the shift amount sW for each combination of the MipSizeld and the modeld, the processing of generating the predicted image of the MIP becomes complicated, or the like.

Therefore, in the third generation method, the shift amount sW of Equation (2) for generating the predicted image of the MIP proposed in Reference A is set to the fixed value.

As the fixed value set in the shift amount sW, for example, among 6 to 8 described in Table 8-8 of Reference A, the minimum value of 6, the maximum value of 8, or the median value of 7 is used, or other values can be adopted. In FIG. 6, for example, among 6 to 8 described in Table 8-8 of Reference A, the minimum value of 6 is adopted as the fixed value set for the shift amount sW.

Then, in the third generation method, the operation of Equation (2) is changed according to the shift amount sW=6 set to the fixed value of 6, and the predicted image of the MIP is generated according to the changed operation.

Specifically, in the third generation method, the shift amount sW is fixed to 6 for each combination of the MipSizeld and the modeld.

Then, according to the fixed shift amount sW=6, the standard weight matrix mWeight [i] [j] and the variable fO (used to obtain the variable oW) corresponding to the bias vector vBias [j] are changed so that the fixed predicted pixel predMip [x] [y] obtained according to Equation (2) using the shift amount sW=6 is a value approximating the standard predicted pixel predMip [x] [y].

Here, the variable fO before setting the shift amount sW to the fixed value, here, the variable fO specified in Reference A is also referred to as the standard variable fO. Further, the variable fO after setting the shift amount sW to the fixed value, here, the variable fO obtained by changing the standard variable fO according to the fixed shift amount sW=6 is also referred to as the fixed variable fO.

In the third generation method, the predicted image of the MIP is generated according to the operation of Equation (2) (and Equation (3)) including the fixed weight matrix mWeight [i] [j] and the fixed variable fO obtained by changing the standard weight matrix mWeight [i] [j] and the standard variable fO according to the shift amount sW=6 set to the fixed value as described above.

As described above, similar to the second generation method, in the third generation method, the MIP is performed using the shift amount sW set to the fixed value. That is, the operation of Equation (2) is changed according to the shift amount sW set to the fixed value, and the predicted image of the MIP is generated according to the changed operation.

Therefore, in the third generation method, the shift amount sW is fixed regardless of the modeld and the combination of the MipSizeld and the modeld, so the processing of the MIP can be simplified. As a result, it is not necessary to specify the table specified as Table 8-8 in Reference A in the standard, and the standard can be simplified.

In addition, for example, when the third generation method is implemented by hardware, the selector for switching the shift amount sW is not required, which can suppress the increases in the circuit scale. Furthermore, when the third generation method is implemented by software, it is not necessary to reference the table specified as Table 8-8 in Reference A, and it is possible to suppress the decrease in the processing speed compared to the case of referencing the table.

FIG. 7 is a diagram for describing a fourth generation method of a predicted image of MIP.

In the fourth generation method, the shift amount sW of the operation for generating the predicted image of the MIP proposed in Reference A is set to the fixed value according to the matrix size of the weight matrix mWeight [i] [j] used in the MIP.

Here, since the MipSizeld is an identifier of the matrix size of the weight matrix mWeight [i] [j], setting the shift amount sW to the fixed value according to the matrix size is equivalent to setting the shift amount sW to the fixed value according to the MipSizeld, for example, setting the shift amount sW to the fixed value for each MipSizeld, or the like.

In the fourth generation method, the operation to generate the predicted image of the MIP is changed according to the shift amount sW set to the fixed value for each MipSizeld, and the predicted image of the MIP is generated according to the changed operation.

In the fourth generation method, as the fixed value for each MipSizeld set in the shift amount sW, for example, the minimum value or the maximum value of the shift amount sW for each MipSizeld described in Table 8-8 of Reference A or other values can be adopted.

For example, as the fixed value for each MipSizeld set in the shift amount sW, the minimum value of the shift amount sW for each MipSizeld described in Table 8-8 of Reference A is adopted.

In Reference A, since the shift amount sW of MipSizeld=0 can be 6 or 7, when the minimum value of the shift amount sW for each MipSizeld described in Table 8-8 of Reference A is adopted as the fixed value for each MipSizeld set in the shift amount sW, in the fourth generation method, the shift amount sW of MipSizeld=0 is fixed to 6 which is the minimum value of 6 and 7. That is, the shift amount sW which is 7 of MipSizeld=0 described in Table 8-8 of Reference A is changed to 6.

In Reference A, since the shift amount sW of MipSizeld=1 can be 7 or 8, when the minimum value of the shift amount sW for each MipSizeld described in Table 8-8 of Reference A is adopted as the fixed value for each MipSizeld set in the shift amount sW, in the fourth generation method, the shift amount sW of MipSizeld=1 is fixed to 7 which is the minimum value of 7 and 8. That is, the shift amount sW which is 8 of MipSizeld=1 described in Table 8-8 of Reference A is changed to 7.

In Reference A, since the shift amount sW of MipSizeld=2 can be 6 or 7, when the minimum value of the shift amount sW for each MipSizeld described in Table 8-8 of Reference A is adopted as the fixed value for each MipSizeld set in the shift amount sW, in the fourth generation method, the shift amount sW of MipSizeld=2 is fixed to 6 which is the minimum value of 6 and 7. That is, the shift amount sW which is 7 of MipSizeld=2 described in Table 8-8 of Reference A is changed to 6.

In the fourth generation method, as described above, the operation of Equation (2) is changed according to the shift amount sW set to the fixed value for each MipSizeld, and the predicted image of the MIP is generated according to the changed operation.

Specifically, in the fourth generation method, the shift amount sW is fixed to 6 for MipSizeld=0 and 2.

Further, for the fixed shift amount sW=6, the standard weight matrix mWeight [i] [j] and the standard variable fO are changed so that the fixed prediction pixel predMip [x] [y] obtained according to Equation (2) using the shift amount sW=6 is the value approximating the standard prediction pixel predMip [x] [y].

Then, the predicted image of the MIP is generated according to the operation of Equation (2) including the fixed weight matrix mWeight [i] [j] and the fixed variable fO obtained by changing the standard weight matrix mWeight [i] [j] and the standard variable fO according to the shift amount sW=6 set to the fixed value.

In addition, for MipSizeld=1, the shift amount sW is fixed to 7.

Furthermore, for the fixed shift amount sW=7, the standard weight matrix mWeight [i] [j] and the standard variable fO are changed so that the fixed prediction pixel predMip [x] [y] obtained according to Equation (2) using the shift amount sW=7 is the value approximating the standard prediction pixel predMip [x] [y].

Therefore, in the fourth generation method, two equations of the equation of the operation in which the operation of Equation (2) is changed according to the shift amount sW fixed to 6, and the equation of the operation in which the operation of Equation (2) is changed according to the shift amount sW fixed to 7 are used to be switched according to the MipSizeld.

As described above, in the fourth generation method, the MIP is performed using the shift amount set to the fixed value for each MipSizeld. That is, the operation of Equation (2) is changed according to the shift amount sW set to the fixed value for each MiPSizeId, and the predicted image of the MIP is generated according to the changed operation.

Therefore, in the fourth generation method, the shift amount sW is fixed regardless of the modeld, so the MIP processing can be simplified, and the table for each modeld specified as Table 8-8 in Reference A is not defined in the standard, and the standard can be simplified.

In addition, for example, when the fourth generation method is implemented by hardware, the selector for switching the shift amount sW is not required, which can suppress the increases in the circuit scale. Furthermore, when the fourth generation method is implemented by software, it is not necessary to reference the table specified as Table 8-8 in Reference A, and it is possible to suppress the decrease in the processing speed compared to the case of referencing the table.

Hereinafter, the shift amount sW described in Reference A, the (standard) weight matrix mWeight [i] [j], the (standard) variable fO, the shift amount sW used in the third generation method and the fourth generation method, the (fixed) weight matrix mWeight [i] [j], and the (fixed) variable fO will be further described.

FIG. 8 is a diagram illustrating the shift amount sW described in Reference A.

In FIG. 8, at the m+1th value from the left, the M+1th value from the top represent the shift amount sW of (M, m). The same applies to the diagram of the shift amount sW described later.

FIG. 9 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 0) described in Reference A.

In FIG. 9, at the i+1th value from the left, the j+1th value from the top represent the standard weight matrix mWeight [i] [j]. The same applies to the diagram of the weight matrix mWeight [i] [j] described later.

FIG. 10 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 1) described in Reference A.

FIG. 11 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 2) described in Reference A.

FIG. 12 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 3) described in Reference A.

FIG. 13 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 4) described in Reference A.

FIG. 14 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 5) described in Reference A.

FIG. 15 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 6) described in Reference A.

FIG. 16 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 7) described in Reference A.

FIG. 17 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 8) described in Reference A.

FIG. 18 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 9) described in Reference A.

FIG. 19 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 10) described in Reference A.

FIG. 20 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 11) described in Reference A.

FIG. 21 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 12) described in Reference A.

FIG. 22 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 13) described in Reference A.

FIG. 23 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 14) described in Reference A.

FIG. 24 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 15) described in Reference A.

FIG. 25 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 16) described in Reference A.

FIG. 26 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(0, 17) described in Reference A.

FIG. 27 is a diagram illustrating the standard variable fO of MipSizeld=0 described in Reference A.

In the standard variable fO of MipSizeld=0 in FIG. 27, the m+1th value from the left represents the standard variable fO of modeld=m. The same applies to the diagram of the variable fO described later.

FIG. 28 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 0) described in Reference A.

FIG. 29 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 1) described in Reference A.

FIG. 30 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 2) described in Reference A.

FIG. 31 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 3) described in Reference A.

FIG. 32 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 4) described in Reference A.

FIG. 33 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 5) described in Reference A.

FIG. 34 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 6) described in Reference A.

FIG. 35 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 7) described in Reference A.

FIG. 36 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 8) described in Reference A.

FIG. 37 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(1, 9) described in Reference A.

FIG. 38 is a diagram illustrating the standard variable fO of MipSizeld=1 described in Reference A.

FIG. 39 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 0) described in Reference A.

FIG. 40 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 1) described in Reference A.

FIG. 41 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 2) described in Reference A.

FIG. 42 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 3) described in Reference A.

FIG. 43 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 4) described in Reference A.

FIG. 44 is a diagram illustrating the standard weight matrix mWeight [i] [j] of (M, m)=(2, 5) described in Reference A.

FIG. 45 is a diagram illustrating the standard variable fO of MipSizeld=2 described in Reference A.

FIG. 46 is a diagram illustrating the shift amount sW used in the third generation method.

In the third generation method, the shift amount sW is set to the fixed value, for example, 6.

Then, in the third generation method, the standard weight matrix mWeight [i] [j], the fixed weight matrix mWeight [i] [j] obtained by changing the standard variable fO, and the fixed variable fO are used so that the fixed predicted pixel predMip [x] [y] obtained after setting the shift amount sW to the fixed value of 6 is a value approximating the standard predicted pixel predMip [x] [y].