Image coding apparatus, image coding method, and storage media
An orthogonal transform unit orthogonally transforms prediction residuals in a block of a P×Q array of pixels, thereby generating an N×M (N is an integer satisfying N<P, and M is an integer satisfying M<Q) array of orthogonal transform coefficients. A quantization unit quantizes the N×M array of the orthogonal transform coefficients using at least a quantization matrix of an N×M array of elements, and thereby generates an N×M array of quantized coefficients.
1 . An image decoding apparatus that decodes, from a bit stream, an image using a plurality of blocks including a first block of a P×Q array of pixels (P and Q are integers), the image decoding apparatus comprising:
a decoding unit configured to decode data corresponding to quantized transform coefficients from the bitstream;
an inverse quantization unit configured to derive transform coefficients from the quantized transform coefficients by using a quantization matrix, the transform coefficients representing frequency components;
an inverse transform unit configured to derive prediction residuals from the transform coefficients by performing inverse transform processing on the transform coefficients; and
a reconstruction unit configured to derive a predicted image by using a prediction method selected from options including at least an intra prediction and a prediction method in which an intra prediction method and an inter prediction method are combined, and to reconstruct image data by using the predicted image and the prediction residuals,
wherein, in a case where a target block is the first block,
(i) the inverse quantization unit is configured to derive an N×M array of transform coefficients from an N×M array of quantized transform coefficients by using a quantization matrix of an N×M array of elements, wherein N is an integer satisfying N<P, and M is an integer satisfying M<Q, and
(ii) the inverse transform unit is configured to derive an N×Q array of intermediate values by performing multiplication of the N×M array of transform coefficients and a matrix of a M×Q array, and to derive a P×Q array of prediction residuals for the first block by performing multiplication of a matrix of a P×N array and the N×Q array of intermediate values, and
wherein, in a case where a target block is a second block smaller than the first block,
(i) the inverse quantization unit is configured to derive an array of transform coefficients for the second block from an array of quantized transform coefficients for the second block by using a quantization matrix for the second block, wherein a size of the array of transform coefficients for the second block, a size of the array of quantized transform coefficients for the second block, and a size of the quantization matrix for the second block are the same as a size of the second block, and
(ii) the inverse transform unit is configured to derive an array of prediction residuals a size of which is the same as the size of the second block by performing multiplication using at least the array of transform coefficients for the second block and a matrix a size of which is the same as the size of the second block.
2 . The image decoding apparatus according to claim 1 , wherein the first block is a square block.
3 . The image decoding apparatus according to claim 2 , wherein the P and the Q are 64, and the N and the M are 32.
4 . The image decoding apparatus according to claim 2 , wherein the P and the Q are 128, and the N and the M are 32.
5 . The image decoding apparatus according to claim 1 , wherein the second block is a block of an N×M array of pixels.
6 . The image decoding apparatus according to claim 1 , wherein the first block is a non-square block.
7 . An image decoding method for decoding, from a bit stream, an image using a plurality of blocks including a first block of a P×Q array of pixels (P and Q are integers), the image decoding method comprising:
decoding data corresponding to quantized transform coefficients from the bitstream;
deriving transform coefficients from the quantized transform coefficients by using a quantization matrix, the transform coefficients representing frequency components;
deriving prediction residuals from the transform coefficients by performing inverse transform processing on the transform coefficients; and
deriving a predicted image by using a prediction method selected from options including at least an intra prediction and a prediction method in which an intra prediction method and an inter prediction method are combined, and reconstructing image data by using the predicted image and the prediction residuals,
wherein, in a case where a target block is the first block,
(i) an N×M array of transform coefficients is derived from an N×M array of quantized transform coefficients by using a quantization matrix of an N×M array of elements, wherein N is an integer satisfying N<P, and M is an integer satisfying M<Q, and
(ii) an N×Q array of intermediate values is derived by performing multiplication of the N×M array of transform coefficients and a matrix of a M×Q array, and a P×Q array of prediction residuals for the first block is derived by performing multiplication of a matrix of a P×N array and the N×Q array of intermediate values, and
wherein, in a case where a target block is a second block smaller than the first block,
(i) an array of transform coefficients for the second block is derived from an array of quantized transform coefficients for the second block by using a quantization matrix for the second block, wherein a size of the array of transform coefficients for the second block, a size of the array of quantized transform coefficients for the second block, and a size of the quantization matrix for the second block are the same as a size of the second block, and
(ii) an array of prediction residuals a size of which is the same as the size of the second block is derived by performing multiplication using at least the array of transform coefficients for the second block and a matrix a size of which is the same as the size of the second block.
8 . A non-transitory computer readable storage medium storing a program for causing a computer to execute an image decoding method for decoding, from a bit stream, an image using a plurality of blocks including a first block of a P×Q array of pixels (P and Q are integers), the image decoding method comprising:
decoding data corresponding to quantized transform coefficients from the bitstream;
deriving transform coefficients from the quantized transform coefficients by using a quantization matrix, the transform coefficients representing frequency components;
deriving prediction residuals from the transform coefficients by performing inverse transform processing on the transform coefficients; and
deriving a predicted image by using a prediction method selected from options including at least an intra prediction and a prediction method in which an intra prediction method and an inter prediction method are combined, and reconstructing image data by using the predicted image and the prediction residuals,
wherein, in a case where a target block is the first block,
(i) an N×M array of transform coefficients is derived from an N×M array of quantized transform coefficients by using a quantization matrix of an N×M array of elements, wherein N is an integer satisfying N<P, and M is an integer satisfying M<Q, and
(ii) an N×Q array of intermediate values is derived by performing multiplication of the N×M array of transform coefficients and a matrix of a M×Q array, and a P×Q array of prediction residuals for the first block is derived by performing multiplication of a matrix of a P×N array and the N×Q array of intermediate values, and
wherein, in a case where a target block is a second block smaller than the first block,
(i) an array of transform coefficients for the second block is derived from an array of quantized transform coefficients for the second block by using a quantization matrix for the second block, wherein a size of the array of transform coefficients for the second block, a size of the array of quantized transform coefficients for the second block, and a size of the quantization matrix for the second block are the same as a size of the second block, and
(ii) an array of prediction residuals a size of which is the same as the size of the second block is derived by performing multiplication using at least the array of transform coefficients for the second block and a matrix a size of which is the same as the size of the second block.