Transform-based image coding method and device
An image decoding method according to the present disclosure includes deriving transform coefficients through dequantization based on the quantized transform coefficients for the target block; deriving modified transform coefficients based on an inverse reduced secondary transform (RST) for the transform coefficients; deriving residual samples for the target block based on an inverse primary transform for the modified transform coefficients; and generating a reconstructed samples based on the residual samples, and prediction samples derived based on an intra prediction mode for the target block, wherein the inverse RST is performed based on a transform kernel matrix selected from a transform set including a plurality of transform kernel matrices, the transform set is determined based on a mapping relationship according to the intra prediction mode applied to the target block, and a plurality of intra prediction modes including the intra prediction mode of the target block are mapped to one transform set.
1. An image decoding method performed by a decoding apparatus, the method comprising:
deriving quantized transform coefficients for a target block from a bitstream;
deriving transform coefficients through dequantization based on the quantized transform coefficients for the target block;
deriving modified transform coefficients based on an inverse non-separable secondary transform for the transform coefficients;
deriving residual samples for the target block based on an inverse primary transform for the modified transform coefficients; and
generating a reconstructed picture based on the residual samples for the target block and prediction samples derived based on an intra prediction mode for the target block,
wherein the inverse non-separable secondary transform derives a larger number of output data than input data through a matrix operation of the input data for a top-left area of the target block and a transform kernel matrix,
wherein the top-left area is a top-left 4×4 area of the target block,
wherein the inverse non-separable secondary transform is performed based on the transform kernel matrix selected from a transform set including a plurality of transform kernel matrices,
wherein the transform set is determined based on a mapping relationship according to the intra prediction mode applied to the target block, and
wherein a plurality of intra prediction modes including the intra prediction mode of the target block are mapped to one transform set.
2. The image decoding method of claim 1 , wherein:
a first intra prediction mode for a first target block to which the inverse non-separable secondary transform is applied and a second intra prediction mode for a second target block to which the inverse non-separable secondary transform is applied are different from each other; and
the transform sets mapped to the first intra prediction mode and the second intra prediction mode are same.
3. The image decoding method of claim 1 , wherein five or more intra prediction modes are mapped to one transform set.
4. The image decoding method of claim 1 , wherein the modified transform coefficients are mapped into the target block based on a diagonal scan order.
5. The image decoding method of claim 1 , further comprising obtaining information on whether the inverse non-separable secondary transform is applied,
wherein the information is included in a sequence parameter set (SPS) in the bitstream.
6. The image decoding method of claim 1 , wherein a matrix coefficient in the transform kernel matrix is expressed in 8 bits.
7. The image decoding method of claim 1 , wherein the inverse non-separable secondary transform is performed based on an inverse 4×4 non-separable secondary transform or an inverse 8×8 non-separable secondary transform determined based on the width and height of the target block, and the inverse 4×4 non-separable secondary transform or the inverse 8×8 non-separable secondary transform is performed based on same transform set.
8. An image encoding method performed by an image encoding apparatus, the method comprising:
deriving prediction samples based on an intra prediction mode applied to a target block;
deriving residual samples for the target block based on the prediction samples;
deriving transform coefficients for the target block based on a primary transform for the residual samples;
deriving modified transform coefficients based on a non-separable secondary transform for the transform coefficients; and
deriving quantized transform coefficients by performing quantization based on the modified transform coefficients,
wherein the non-separable secondary transform derives a smaller number of output data than input data through a matrix operation of the input data for a top-left area of the target block and a transform kernel matrix,
wherein the top-left area is a top-left 4×4 area of the target block or a top-left 8×8 area of the target block,
wherein the non-separable secondary transform is performed based on the transform kernel matrix selected from a transform set including a plurality of transform kernel matrices,
wherein the transform set is determined based on a mapping relationship according to the intra prediction mode applied to the target block, and
wherein a plurality of intra prediction modes including the intra prediction mode of the target block are mapped to one transform set.
9. The image encoding method of claim 8 , wherein:
the target block includes a first block and a second block;
a first intra prediction mode for a first target block to which the inverse non-separable secondary transform is applied and a second intra prediction mode for a second target block to which the inverse non-separable secondary transform is applied are different from each other; and
and the transform sets mapped to the first intra prediction mode and the second intra prediction mode are same.
10. The image encoding method of claim 8 , wherein five or more intra prediction modes are mapped to one transform set.
11. The image encoding method of claim 8 , further comprising encoding information on whether the inverse non-separable secondary transform is applied,
wherein the information is included in a sequence parameter set (SPS) in the bitstream.
12. The image encoding method of claim 8 , wherein the non-separable secondary transform is performed based on a 4×4 non-separable secondary transform or an 8×8 non-separable secondary transform determined based on the width and height of the target block, and the 4×4 non-separable secondary transform or the 8×8 non-separable secondary transform is performed based on same transform set.
13. A non-transitory computer readable storage medium storing a bitstream generated by a method, the method comprising:
deriving prediction samples based on an intra prediction mode applied to a target block;
deriving residual samples for the target block based on the prediction samples;
deriving transform coefficients for the target block based on a primary transform for the residual samples;
deriving modified transform coefficients based on a non-separable secondary transform for the transform coefficients; and
deriving quantized transform coefficients by performing quantization based on the modified transform coefficients, and
encoding image information on the quantized transform coefficients to generate the bitstream,
wherein the non-separable secondary transform derives a smaller number of output data than input data through a matrix operation of the input data for a top-left area of the target block and a transform kernel matrix,
wherein the top-left area is a top-left 4×4 area of the target block or a top-left 8×8 area of the target block,
wherein the inverse non-separable secondary transform RST is performed based on the transform kernel matrix selected from a transform set including a plurality of transform kernel matrices,
wherein the transform set is determined based on a mapping relationship according to the intra prediction mode applied to the target block, and
wherein a plurality of intra prediction modes including the intra prediction mode of the target block are mapped to one transform set.
14. The image decoding method of claim 1 , wherein a range of values of matrix coefficients in the transform kernel matrix is from −128 to 127.
15. The image encoding method of claim 8 , wherein a range of values of matrix coefficients in the transform kernel matrix is from −128 to 127.
16. The non-transitory computer readable storage medium of claim 13 , wherein a range of values of matrix coefficients in the transform kernel matrix is from −128 to 127.