Sample mapping for low dynamic range-based image and video coding

1 . A method for decoding a block in a bitstream of a video or an image, comprising:

receiving the bitstream;

determining, from the bitstream, that a Low Dynamic Range (LDR) is applied to encoding the block;

decoding at least one color component of the block to generate reconstructed samples of the block at a first bit depth N 1 ;

determining an LDR sample mapping scheme; and

applying the LDR sample mapping scheme to the reconstructed samples of the block to generate mapped samples of the block having a second bit depth N 2 , the LDR sample mapping scheme being conditionally determined among a plurality of LDR sample mapping schemes dependent on coded information comprising one or more of: reconstructed color component samples of the block, reconstructed neighboring sample values the block, quantization parameters of the block, frame resolution associated with the block, or temporal layer information associated with the block, and N 2 being larger than N 1 .

2 . The method of claim 1 , wherein the LDR sample mapping scheme comprises mapping the reconstructed samples using a lookup table to generate the mapped samples, the lookup table being predefined or being signaled in the bitstream.

3 . The method of claim 1 , wherein the LDR sample mapping scheme comprises mapping the reconstructed samples using a mapping function, the mapping function being predefined or being signaled in the bitstream.

4 . The method of claim 3 , wherein the mapping function comprises an algorithmic operation comprising:

an operation of left shift of N 2 −N 1 bits; or

a left shift operation followed by adding an offset value; or

a Dynamic Range Expansion (DRE).

5 . The method of claim 3 , wherein the mapping function comprises a trained deep learning model, the trained deep learning model comprising at least one of a Convolutional Neural Network (CNN), a Generative Adversarial Network (GAN), or a diffusion model.

6 . The method of claim 3 , wherein the mapping function comprises both an algorithmic mapping function and a trained deep learning model.

7 . The method of claim 1 , wherein determining the LDR sample mapping scheme comprises decoding the bitstream to obtain an explicitly signaled syntax element for indicating the LDR sample mapping scheme.

8 . The method of claim 7 , wherein the explicitly signaled syntax element in the bitstream is obtained from:

a high-level syntax comprising a sequence header, a picture header, a subpicture header, a frame header, a slice header, or a tile header; or

a Supplemental Enhancement Information (SEI) message.

9 . The method of claim 7 , wherein the explicitly signaled syntax element in the bitstream is signaled for per LDR image area.

10 . The method of claim 1 , wherein the LDR sample mapping scheme comprises mapping the reconstructed samples using a mapping function and wherein:

the mapping function is signaled via an index to a plurality of predefined mapping functions.

11 . The method of claim 1 , further comprising performing a clipping of the mapped samples of the block.

12 . The method of claim 1 , wherein different LDR sample mapping schemes are applied to different color components of the block.

13 . A method for encoding a block a video or an image to generate a bitstream, comprising:

determining that the block is to be encoded by applying an LDR;

determining an encoded bit depth of encoded samples of the block, N 1 ;

determining an original bit depth of original samples of the block, N 2 , N 2 being larger than N 1 ;

applying an LDR sample mapping scheme to the original samples of the block having the original bit depth N 2 to generate mapped LDR samples of the block having the encoded bit depth N 1 , the LDR sample mapping scheme being conditionally determined among a plurality of LDR sample mapping schemes dependent on coded information comprising one or more of: reconstructed color component samples of the block, reconstructed neighboring sample values the block, quantization parameters of the block, frame resolution associated with the block, or temporal layer information associated with the block; and

encoding the mapped LDR samples of the block to generate the encoded samples of the block in the bitstream.

14 . The method of claim 13 , wherein the LDR sample mapping scheme comprises mapping original samples of the block using a lookup table to generate the mapped LDR samples, the lookup table being predefined or being signaled in the bitstream.

15 . The method of claim 13 , wherein the LDR sample mapping scheme comprises mapping the original samples of the block using a mapping function, the mapping function being predefined or being signaled in the bitstream.

16 . The method of claim 15 , wherein the mapping function comprises an algorithmic operation comprising:

an operation of right shift of N 2 −N 1 bits; or

a right shift operation followed by adding an offset value; or

a reverse DRE.

17 . The method of claim 15 , wherein the mapping function comprises a trained deep learning model, the trained deep learning model comprising at least one of a Convolutional Neural Network (CNN), a Generative Adversarial Network (GAN), or a diffusion model.

18 . The method of claim 13 , further comprising including at least one syntax element for indicating the LDR sample mapping scheme in the bitstream.

19 . A method for processing a block of a video or image, comprising converting the block to a bitstream, wherein the bitstream comprises:

an encoded block generated by:

determining an encoded bit depth of encoded samples of the block, N 1 ;

determining an original bit depth of original samples of the block, N 2 , N 2 being larger than N 1 ;

applying an LDR sample mapping scheme to the original samples of the block having the original bit depth N 2 to generate mapped LDR samples of the block having the encoded bit depth N 1 , the LDR sample mapping scheme being conditionally determined among a plurality of LDR sample mapping schemes dependent on coded information comprising one or more of: reconstructed color component samples of the block, reconstructed neighboring sample values the block, quantization parameters of the block, frame resolution associated with the block, or temporal layer information associated with the block; and

encoding the mapped LDR samples of the block to generate the encoded samples of the block in the bitstream; and

at least one syntax element for indicating that the LDR sample mapping scheme is applied or for indicating the LDR sample mapping scheme.