Image coding method based on transform, and device therefor
An image decoding method according to the present document comprises a step of deriving modified transform coefficients, wherein the step of deriving the modified transform coefficients comprises a step of setting an LFNST apply variable about whether to apply an LFNST on the basis of the tree type of the current block, a transform coefficient coding flag for each component of the current block, or a transform skip flag for each component of the current block, and can comprise a step of parsing an LFNST index on the basis of the value of the LFNST apply variable.
1. An image decoding method performed by a decoding apparatus, comprising:
receiving residual information from a bitstream;
deriving transform coefficients for a first component of a current block by performing dequantization based on the residual information, the first component being representative of one of a luma component or a chroma component of the current block; and
deriving modified transform coefficients by applying a low-frequency non-separable transform (LFNST) to the transform coefficients,
wherein deriving the modified transform coefficients comprises:
setting an LFNST apply variable based on a tree type of the current block, a transform skip flag for the first component of the current block, and a transform coefficient coding flag for a second component of the current block, the second component being representative of the other of the luma component or the chroma component of the current block;
parsing, from the bitstream, an LFNST index based on the LFNST apply variable, the LFNST index indicating one of a plurality of LFNST matrices in an LFNST set; and
performing the LFNST on the transform coefficients based on the LFNST index,
wherein a value of the LFNST apply variable is set equal to 1 based on the tree type of the current block being a single tree, the transform skip flag for the first component being equal to 0, and the transform coefficient coding flag for the second component being equal to 0, and
wherein the LFNST index is parsed from the bitstream based on the value of the LFNST apply flag being equal to 1, and
wherein the first component of the current block is not coded in a block-based delta pulse code modulation (BDPCM) mode while the second component of the current block is coded in the BDPCM mode.
2. The image decoding method of claim 1 , wherein the value of the LFNST apply variable is set equal to 1 regardless of a transform coefficient coding flag for the first component of the current block.
3. The image decoding method of claim 2 , wherein a transform skip flag for the second component of the current block is inferred to 1.
4. An image encoding method performed by an encoding apparatus, comprising:
deriving prediction samples for a first component of a current block, the first component being representative of one of a luma component or a chroma component of the current block;
deriving residual samples for the first component of the current block based on the prediction samples;
deriving transform coefficients for the first component of the current block based on a primary transform of the residual samples; and
deriving modified transform coefficients from the transform coefficients by applying a low-frequency non-separable transform (LFNST),
wherein deriving the modified transform coefficients comprises:
setting an LFNST apply variable based on a tree type of the current block, a transform skip flag for the first component of the current block, and a transform coefficient coding flag for a second component of the current block, the second component being representative of the other of the luma component or the chroma component of the current block; and
applying the LFNST based on the LFNST apply variable,
wherein an LFNST index indicating one of a plurality of LFNST matrices in an LFNST set is encoded into a bitstream based on the LFNST apply variable,
wherein a value of the LFNST apply variable is set equal to 1 based on the tree type of the current block being a single tree, the transform skip flag for the first component being equal to 0, and the transform coefficient coding flag for the second component being equal to 0,
wherein the LFNST is applied based on the value of the LFNST apply variable being equal to 1, and
wherein the first component of the current block is not coded in a block-based delta pulse code modulation (BDPCM) mode while the second component of the current block is coded in the BDPCM mode.
5. The image encoding method of claim 4 , wherein the value of the LFNST apply variable is set equal to 1 regardless of a transform coefficient coding flag for the first component of the current block.
6. The image encoding method of claim 5 , wherein a transform skip flag for the second component of the current block is inferred to 1.
7. A non-transitory computer-readable digital storage medium for storing a bitstream generated by an image encoding method,
the image encoding method comprising:
deriving prediction samples for a first component of a current block, the first component being representative of one of a luma component or a chroma component of the current block;
deriving residual samples for the first component of the current block based on the prediction samples;
deriving transform coefficients for the first component of the current block based on a primary transform of the residual samples; and
deriving modified transform coefficients from the transform coefficients by applying a low-frequency non-separable transform (LFNST),
wherein deriving the modified transform coefficients comprises:
setting an LFNST apply variable based on a tree type of the current block, a transform skip flag for the first component of the current block, and a transform coefficient coding flag for a second component of the current block, the second component being representative of the other of the luma component or the chroma component of the current block; and
applying the LFNST based on the LFNST apply variable,
wherein an LFNST index indicating one of a plurality of LFNST matrices in an LFNST set is encoded into the bitstream based on the LFNST apply variable,
wherein a value of the LFNST apply variable is set equal to 1 based on the tree type of the current block being a single tree, the transform skip flag for the first component being equal to 0, and the transform coefficient coding flag for the second component being equal to 0,
wherein the LFNST is applied based on the value of the LFNST apply variable being equal to 1, and
wherein the first component of the current block is not coded in a block-based delta pulse code modulation (BDPCM) mode while the second component of the current block is coded in the BDPCM mode.
8. A method for transmitting data for image information, comprising:
deriving prediction samples for a first component of a current block, the first component being representative of one of a luma component or a chroma component of the current block;
deriving residual samples for the first component of the current block based on the prediction samples;
deriving transform coefficients for the first component of the current block based on a primary transform of the residual samples;
deriving modified transform coefficients from the transform coefficients by applying a low-frequency non-separable transform (LFNST);
deriving quantized transform coefficients by performing a quantization on the modified transform coefficients;
encoding residual information for the quantized transform coefficients to generate a bitstream; and
transmitting the data comprising the bitstream,
wherein deriving the modified transform coefficients comprises:
setting an LFNST apply variable based on a tree type of the current block, a transform skip flag for the first component of the current block, and a transform coefficient coding flag for a second component of the current block, the second component being representative of the other of the luma component or the chroma component of the current block; and
applying the LFNST based on the LFNST apply variable,
wherein an LFNST index indicating one of a plurality of LFNST matrices in an LFNST set is encoded into the bitstream based on the LFNST apply variable,
wherein a value of the LFNST apply variable is set equal to 1 based on the tree type of the current block being a single tree, the transform skip flag for the first component being equal to 0, and the transform coefficient coding flag for the second component being equal to 0,
wherein the LFNST is applied based on the value of the LFNST apply variable being equal to 1, and
wherein the first component of the current block is not coded in a block-based delta pulse code modulation (BDPCM) mode while the second component of the current block is coded in the BDPCM mode.