Cross-channel prediction based on multiple prediction modes
This disclosure describes a set of advanced video coding technologies. More specifically, the disclosed technology involves cross-channel prediction and particularly chroma from luma (CfL) prediction. In some example implementations, a sample in a chroma channel is predicted based on a weighted sum or average of an alternate current prediction contribution derived based on reconstructed samples of a luma channel and a direct current prediction contribution generated based on an intra-prediction of the sample in the chroma channel based on reconstructed samples of the chroma channel, the intra-prediction being performed based on an intra-prediction mode other than a DC mode.
1. A method for video processing, the method comprising:
determining that a chroma from luma (CfL) prediction mode is to be applied to a current sample of a chroma channel in a current coding block based on at least one reconstructed sample from a luma channel in a received bitstream;
deriving an alternating current (AC) contribution to a predictor of the current sample of the chroma channel based on the at least one reconstructed sample from the luma channel;
deriving a direct current (DC) contribution to the predictor of the current sample of the chroma channel by performing an intra-prediction of the current sample of the chroma channel based on at least one reconstructed sample of the chroma channel using one of a subset of intra-prediction modes that exclude a DC intra-prediction mode, wherein the DC intra-prediction mode involves an average of a reconstructed above neighboring sample of the current sample in an above neighboring block and a reconstructed left neighboring sample of the current sample in an left neighboring block;
computing a weighted average of the AC contribution and the derived DC contribution to generate the predictor of the current sample of the chroma channel; and
reconstructing the current sample of the chroma channel based on at least the predictor of the current sample of the chroma channel.
2. The method of claim 1 , further comprising subsampling the luma channel of the current coding block to obtain a subsampled luma block corresponding to a chroma block in the chroma channel of the current coding block.
3. The method of claim 2 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel comprises one of SMOOTH, SMOOTH-H, SMOOTH-V, PAETH, directional, recursive filtering, or offset-based refinement intra-prediction modes.
4. The method of claim 3 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel is predetermined.
5. The method of claim 3 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel is selected from a predetermined subset of intra-prediction modes according to explicit signaling in the received bitstream.
6. The method of claim 3 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel is derived based on a reference intra-prediction mode associated with a luma sample in the subsampled luma block, the luma sample being collocated with the current sample of the chroma channel.
7. The method of claim 3 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel is derived from a reference intra-prediction mode associated with a neighboring chroma block of the current coding block.
8. The method of claim 3 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel is derived according to a size of the current coding block.
9. The method of claim 3 , wherein deriving the AC contribution to the predictor of the current sample of the chroma channel comprises subtracting a sample average of the subsampled and reconstructed luma block from a reconstructed luma sample collocated with the current sample of the chroma channel.
10. The method of claim 3 , wherein deriving the AC contribution to the predictor of the current sample of the chroma channel comprises subtracting a prediction of a luma sample collocated with the current sample of the chroma channel from the luma sample as reconstructed.
11. The method of claim 10 , wherein the prediction of the luma sample is generated under an intra-prediction mode.
12. The method of claim 11 , wherein the intra prediction mode for generating the prediction of the luma sample comprises one of SMOOTH, SMOOTH-H, SMOOTH-V, PAETH, directional, recursive filtering, or offset-based refinement intra-prediction modes.
13. The method of claim 12 , wherein the intra prediction mode for generating the prediction of the luma sample is explicitly signaled in the received bitstream.
14. The method of claim 11 , wherein the intra prediction mode for generating the prediction of the luma sample is implicitly derived.
15. The method of claim 14 , wherein the intra prediction mode for generating the prediction of the luma sample is implicitly derived as the same as the intra-prediction mode for deriving the DC contribution to the predictor of the current sample of the chroma channel.
16. The method of claim 14 , wherein the intra prediction mode for generating the prediction of the luma sample is derived as a reference intra-prediction mode applied to the luma sample collocated with the current sample in the chroma channel.
17. The method of claim 14 , wherein the intra prediction mode for generating the prediction of the luma sample is derived from a reference intra-prediction mode associated with a neighboring luma block of the current coding block.
18. An electronic device for video processing, comprising a memory for storing instructions and a processor for executing the instructions to cause the electronic device to:
determine that a cross-channel prediction mode is to be applied to a current sample of a chroma channel in a current coding block based on at least one reconstructed sample from a luma channel in a received bitstream;
derive an alternating current (AC) contribution to a predictor of the current sample of the chroma channel based on the at least one reconstructed sample from the luma channel;
derive a direct current (DC) contribution to the predictor of the current sample of the chroma channel by performing an intra-prediction of the current sample of the chroma channel based on at least one reconstructed sample of the chroma channel using one of a subset of intra-prediction mode that exclude a DC intra-prediction mode, wherein the DC intra-prediction mode involves an average of a reconstructed above neighboring sample of the current sample in an above neighboring block and a reconstructed left neighboring sample of the current sample in an left neighboring block;
compute a weighted average of the AC contribution and the derived DC contribution to generate the predictor of the current sample of the chroma channel; and
reconstruct the current sample of the chroma channel based on at least the predictor of the current sample of the chroma channel.
19. The electronic device of claim 18 , wherein the intra-prediction mode for performing the intra-prediction to derive the DC contribution to the predictor of the current sample of the chroma channel comprises one of SMOOTH, SMOOTH-H, SMOOTH-V, PAETH, directional, recursive filtering, or offset-based refinement intra-prediction modes.
20. A method for video processing, the method comprising:
determining that a chroma from luma (CfL) prediction mode is to be applied to a current sample of a chroma channel in a current coding block based on at least one reconstructed sample from a luma channel in a received bitstream;
deriving an alternating current (AC) contribution to a predictor of the current sample of the chroma channel based on the at least one reconstructed sample from the luma channel;
deriving a direct current (DC) contribution to the predictor of the current sample of the chroma channel by performing an intra-prediction of the current sample of the chroma channel based on at least one reconstructed sample of the chroma channel using one of a subset of intra-prediction modes that exclude a DC intra-prediction mode, wherein the DC intra-prediction mode involves an average of a reconstructed above neighboring sample of the current sample in an above neighboring block and a reconstructed left neighboring sample of the current sample in an left neighboring block;
computing a weighted average of the AC contribution and the DC contribution to generate the predictor of the current sample of the chroma channel; and
encoding the current sample of the chroma channel based on the predictor.