Video encoding apparatus and video decoding apparatus and encoding method and decoding method thereof
A video encoding apparatus for encoding a plurality of image blocks in a video frame includes an intra-frame prediction module, a transformation module and a quantization module. The intra-frame prediction module performs intra-frame prediction on the image blocks to generate a plurality of residual blocks. The transformation module performs a transformation on a target residual block along a predetermined direction according to a transformation matrix to generate a transformation result. The transformation matrix is a product of an initial transformation matrix and a secondary transformation matrix. The initial transformation matrix corresponds to a one-dimensional initial transform performed along the predetermined direction in a two-dimensional initial transform. The secondary transformation matrix corresponds to a one-dimensional secondary transform performed along the predetermined direction in a two-dimensional secondary transform. The quantization module quantizes the transformation result.
1. A video encoding apparatus, for encoding a plurality of image blocks in a video frame, the video encoding apparatus comprising:
an intra-frame prediction module, configured to perform intra-frame prediction on the plurality of image blocks to generate a plurality of residual blocks;
a transformation module, configured to perform a transform on a target residual block along a predetermined direction according to a transformation matrix to generate a transformation result, wherein the transformation matrix is based on arranging a horizontal row secondary transform between a horizontal row initial transform and a vertical column initial transform;
wherein, the transformation matrix is a product of an initial transformation matrix and a secondary transformation matrix, the initial transformation matrix corresponds to a one-dimensional initial transform performed along the predetermined direction in a two-dimensional initial transform, and the secondary transformation matrix corresponds to a one-dimensional secondary transform performed along the predetermined direction in a two-dimensional secondary transform, wherein the one-dimensional initial transform is a discrete cosine transform performed along the predetermined direction in the two-dimensional initial transform, and the one-dimensional secondary transform is performed on low-frequency components in the initial transformation matrix; and
a quantization module, configured to quantize the transformation result.
2. The video encoding apparatus according to claim 1 , wherein the transformation module is further configured to perform another transform perpendicular to the predetermined direction on the target residual block according to another transformation matrix; the another transformation matrix is a product of another initial transformation matrix and another secondary transformation matrix, the another initial transformation matrix corresponds to a one-dimensional initial transform perpendicular to the predetermined direction in the two-dimensional initial transform, and the another secondary transformation matrix corresponds to a one-dimensional secondary transform perpendicular to the predetermined direction in the two-dimensional secondary transform.
3. The video encoding apparatus according to claim 1 , wherein sizes of the two-dimensional initial transform and the two-dimensional secondary transform are the same.
4. The video encoding apparatus according to claim 1 , wherein the transformation matrix is:
[
34
77
79
55
58
69
-
33
-
84
72
-
7
-
75
73
81
-
75
58
-
28
]
.
5. The video encoding apparatus according to claim 1 , being applied to a digital audio video standard (AVS) encoding system.
6. The video encoding apparatus according to claim 1 , wherein when the target residual block does not require the one-dimensional secondary transform performed along the predetermined direction, the transformation module sets the secondary transformation matrix as a unit matrix.
7. The video encoding apparatus according to claim 1 , wherein a size of the two-dimensional initial transform perpendicular to the predetermined direction is greater than a size of the two-dimensional secondary transform perpendicular to the predetermined direction; the transformation module is further configured to:
before performing the transform, perform a one-dimensional initial transform perpendicular to the predetermined direction in the two-dimensional initial transform on an input image data block to generate an initial transformation result comprising a first sub-block and a second sub-block; wherein the first sub-block is the target residual block, and size of the first sub-block perpendicular to the predetermined direction is equal to the size of the two-dimensional secondary transform perpendicular to the predetermined direction;
perform the one-dimensional initial transform along the predetermined direction in the two-dimensional initial transform on the second sub-block; and
perform a one-dimensional secondary transform perpendicular to the predetermined direction in the two-dimensional secondary transform on an intermediate transform result generated from performing the transform on the target residual block.
8. A video encoding method, comprising:
performing a transform on a target residual block along a predetermined direction according to a transformation matrix, wherein the transformation matrix is based on arranging a horizontal row secondary transform between a horizontal row initial transform and a vertical column initial transform;
wherein, the transformation matrix is a product of an initial transformation matrix and a secondary transformation matrix, the initial transformation matrix corresponds to a one-dimensional initial transform performed along the predetermined direction in a two-dimensional initial transform, and the secondary transformation matrix corresponds to a one-dimensional secondary transform performed along the predetermined direction in a two-dimensional secondary transform, wherein the one-dimensional initial transform is a discrete cosine transform performed along the predetermined direction in the two-dimensional initial transform, and the one-dimensional secondary transform is performed on low-frequency components in the initial transformation matrix.
9. The video encoding method according to claim 8 , further comprising:
performing another transform perpendicular to the predetermined direction on the target residual block according to another transformation matrix;
wherein, the another transformation matrix is a product of another initial transformation matrix and another secondary transformation matrix, the another initial transformation matrix corresponds to a one-dimensional initial transform perpendicular to the predetermined direction in the two-dimensional initial transform, and the another secondary transformation matrix corresponds to a one-dimensional secondary transform perpendicular to the predetermined direction in the two-dimensional secondary transform.
10. The video encoding method according to claim 8 , wherein sizes of the two-dimensional initial transform and the two-dimensional secondary transform are the same.
11. The video encoding method according to claim 10 , applied to the digital AVS, wherein the transformation matrix is:
[
34
77
79
55
58
69
-
33
-
84
72
-
7
-
75
73
81
-
75
58
-
28
]
.
12. The video encoding method according to claim 8 , wherein when the target residual block does not require the one-dimensional secondary transform performed along the predetermined direction, the secondary transformation matrix is set to a unit matrix.
13. The video encoding method according to claim 8 , wherein a size of the two-dimensional initial transform perpendicular to the predetermined direction is greater than a size of the two-dimensional secondary transform perpendicular to the predetermined direction; the video encoding method further comprising:
before performing the transform, performing a one-dimensional initial transform perpendicular to the predetermined direction in the two-dimensional initial transform on an input image data block to generate an initial transformation result comprising a first sub-block and a second sub-block; wherein the first sub-block is the target residual block, and size of the first sub-block perpendicular to the predetermined direction is equal to the size of the two-dimensional secondary transform perpendicular to the predetermined direction;
performing the one-dimensional initial transform along the predetermined direction in the two-dimensional initial transform on the second sub-block; and
performing a one-dimensional secondary transform perpendicular to the predetermined direction in the two-dimensional secondary transform on an intermediate transform result generated from performing the transform on the target residual block.
14. A video decoding method, decoding a plurality of coded data, the method comprising:
decoding a set of target coded data to generate quantized residual block information;
de-quantizing the quantized residual block information to generate intermediate information;
performing a reverse transform on the intermediate information along a predetermined direction according to a reverse transformation matrix to generate a residual block, wherein the transformation matrix is based on arranging a horizontal row secondary transform between a horizontal row initial transform and a vertical column initial transform;
wherein, the reverse transformation matrix is a product of a reverse initial transformation matrix and a reverse secondary transformation matrix, the reverse initial transformation matrix corresponds to a one-dimensional reverse initial transform performed along the predetermined direction in a two-dimensional reverse initial transform, and the reverse secondary transformation matrix corresponds to a one-dimensional reverse secondary transform performed along the predetermined direction in a two-dimensional reverse secondary transform, wherein the one-dimensional reverse initial transform is a discrete cosine transform performed along the predetermined direction in the two-dimensional initial transform, and the one-dimensional reverse secondary transform is performed on low-frequency components in the reverse initial transformation matrix;
performing an intra-frame prediction on the residual block to generate a prediction block; and
generating a target image block using the prediction block and the residual block.
15. The video decoding method according to claim 14 , further comprising:
performing another reverse transform along a direction perpendicular to the predetermined direction according to another reverse transformation matrix;
wherein, the another reverse transformation matrix is a product of another reverse initial transformation matrix and another reverse secondary transformation matrix, the another reverse initial transformation matrix corresponds to a one-dimensional reverse initial transform perpendicular to the predetermined direction in the two-dimensional initial transform, and the another reverse secondary transformation matrix corresponds to a one-dimensional reverse secondary transform perpendicular to the predetermined direction in the two-dimensional reverse secondary transform.
16. The video decoding method according to claim 14 , wherein sizes of the two-dimensional reverse initial transform and the two-dimensional reverse secondary transform are the same.
17. The video decoding method according to claim 16 , applied to the digital AVS, wherein the predetermined direction is a horizontal direction, and the reverse transformation matrix is:
[
34
58
72
81
77
69
-
7
-
75
79
-
33
-
75
58
55
-
84
73
-
28
]
.
18. The video decoding method according to claim 14 , wherein when it is determined that the one-dimensional reverse secondary transform performed along the predetermined direction is not required, the reverse secondary transformation matrix is set to a unit matrix.
19. The video decoding method according to claim 14 , wherein a size of the two-dimensional reverse initial transform perpendicular to the predetermined direction is greater than a size of the two-dimensional reverse secondary transform perpendicular to the predetermined direction; the video decoding method further comprising:
before performing the transform, performing a one-dimensional reverse initial transform perpendicular to the predetermined direction in the two-dimensional reverse initial transform on an input data to generate an initial transformation result comprising a first sub-block and a second sub-block; wherein the first sub-block is the intermediate information, and a size of the first sub-block perpendicular to the predetermined direction is the size of the two-dimensional reverse secondary transform perpendicular to the predetermined direction;
performing a one-dimensional reverse initial transform along the predetermined direction in the two-dimensional reverse initial transform on the second sub-block; and
performing a one-dimensional reverse secondary transform perpendicular to the predetermined direction in the two-dimensional reverse secondary transform on an intermediate transform result generated from performing the combined reverse transform on the intermediate information.
20. A video encoding apparatus, for encoding a plurality of image blocks in a video frame, the video encoding apparatus comprising:
an intra-frame prediction module comprising a multiplexer to receive, configured to perform intra-frame prediction on the plurality of image blocks to generate a plurality of residual blocks;
a transformation module, configured to perform a transform on a target residual block along a predetermined direction according to a transformation matrix to generate a transformation result;
wherein, the transformation matrix is a product of an initial transformation matrix and a secondary transformation matrix, the initial transformation matrix corresponds to a one-dimensional initial transform performed along the predetermined direction in a two-dimensional initial transform, and the secondary transformation matrix corresponds to a one-dimensional secondary transform performed along the predetermined direction in a two-dimensional secondary transform; and
a quantization module, configured to quantize the transformation result
wherein a size of the two-dimensional initial transform perpendicular to the predetermined direction is greater than a size of the two-dimensional secondary transform perpendicular to the predetermined direction; the transformation module is further configured to:
before performing the transform, perform a one-dimensional initial transform perpendicular to the predetermined direction in the two-dimensional initial transform on an input image data block to generate an initial transformation result comprising a first sub-block and a second sub-block; wherein the first sub-block is the target residual block, and size of the first sub-block perpendicular to the predetermined direction is equal to the size of the two-dimensional secondary transform perpendicular to the predetermined direction;
perform the one-dimensional initial transform along the predetermined direction in the two-dimensional initial transform on the second sub-block; and
perform a one-dimensional secondary transform perpendicular to the predetermined direction in the two-dimensional secondary transform on an intermediate transform result generated from performing the transform on the target residual block.