Methods and apparatus for multi-hypothesis mode reference and constraints
The techniques described herein relate to methods, apparatus, and computer readable media configured to decode and/or encode video data. If the current coding block was using more than one hypotheses of prediction data to generate the prediction data, first prediction data is determined for a current coding block of a picture generated using a first prediction mode, and second prediction data is determined for the current coding block generated using a second prediction mode. Information derived from at least one of the first prediction mode and the second prediction mode can be stored. The stored information can be referenced by the subsequent coding blocks.
1. A decoding method for decoding video data, the method comprising:
decoding a current coding block of a picture of the video data, comprising:
determining whether both inter prediction and intra prediction are used for generating a single set of combined prediction samples used to reconstruct the current coding block of the picture;
upon determining that both inter prediction and intra prediction are used for generating the single set of combined prediction samples of the current coding block of the picture:
determining, based on at least one reconstructed reference frame, first prediction samples for the current coding block generated using a first prediction mode, wherein the first prediction mode comprises an inter-prediction merge mode;
determining, based on at least one spatial neighboring block to the current coding block, second prediction samples for the current coding block generated using a second prediction mode, wherein the second prediction mode comprises an intra prediction mode;
combining the first prediction samples and the second prediction samples to generate the single set of combined prediction samples for the current coding block;
storing first information derived from the first prediction mode, second information referring to the second prediction mode, or both; and
reconstructing the current coding block based on the single set of combined prediction samples for the current coding block of the picture.
2. The decoding method of claim 1 , wherein a second coding block that is a subsequent coding block to the current coding block uses the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
3. The decoding method of claim 1 , further comprising generating a list of most probable modes used in intra prediction for a subsequent coding block to the current coding block according to the second prediction mode.
4. The decoding method of claim 3 , further comprising decoding the subsequent coding block using intra prediction with second information derived from prediction data determined based on at least one spatial neighboring block to the current coding block as a candidate.
5. The decoding method of claim 1 , wherein the at least one spatial neighboring block to the current coding block is a previous coding block.
6. The decoding method of claim 5 , wherein the current coding block and the previous coding block belong to a same coding tree unit (CTU), a same frame, or both.
7. The decoding method of claim 1 , wherein only when subsequent coding blocks and the current coding block belong to the same CTU, the same frame, or both, the subsequent coding blocks uses at least part of the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
8. The decoding method of claim 1 , wherein when subsequent coding blocks and the current coding block belong to different CTUs, different frames, or both, the subsequent coding blocks uses information excluding at least part of the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
9. The decoding method of claim 1 , wherein the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both, is used for a subsequent coding block to reference.
10. The decoding method of claim 1 , further comprising:
combining the first prediction samples and the second prediction samples using a weighted sum to generate the single set of combined prediction samples.
11. An apparatus configured to decode video data, the apparatus comprising a processor in communication with memory, the processor being configured to execute instructions stored in the memory that cause the processor to:
decode a current coding block of a picture of the video data, comprising:
determining whether both inter prediction and intra prediction are used for generating a single set of combined prediction samples used to reconstruct the current coding block of the picture;
upon determining that both inter prediction and intra prediction are used for generating the single set of combined prediction samples of the current coding block of the picture:
determine, based on at least one reconstructed reference frame, first prediction samples for the current coding block generated using a first prediction mode, wherein the first prediction mode comprises an inter-prediction merge mode;
determine, based on at least one spatial neighboring block to the current coding block, second prediction samples for the current coding block generated using a second prediction mode, wherein the second prediction mode comprises an intra prediction mode;
combining the first prediction samples and the second prediction samples to generate the singe set of combined prediction samples;
storing first information derived from the first prediction mode, second information referring to the second prediction mode, or both; and
reconstructing the current coding block based on the single set of combined prediction samples for the current coding block of the picture.
12. An encoding method for encoding video data, the method comprising:
encoding a current coding block of a picture of the video data, comprising:
determining whether both inter prediction and intra prediction are used to generate a single set of combined prediction samples to encode the current coding block of the picture;
upon determining that both inter prediction and intra prediction are used for generating the single set of combined prediction samples of the current coding block of the picture:
generating, based on at least one reference frame, first prediction samples for the current coding block using a first prediction mode, wherein the first prediction mode comprises an inter-prediction merge mode;
generating, based on at least one spatial neighboring block to the current coding block, second prediction samples for the current coding block using a second prediction mode, wherein the second prediction mode comprises an intra prediction mode;
combining the first prediction samples and the second prediction samples to generate the single set of combined prediction samples for the current coding block;
storing first information derived from the first prediction mode, second information referring to the second prediction mode, or both; and
encoding the current coding block based on the single set of combined prediction samples for the current coding block of the picture.
13. The encoding method of claim 12 , further comprising encoding a second coding block that is a subsequent coding block to the current coding block based on the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
14. The encoding method of claim 12 , further comprising:
encoding a subsequent coding block based on the first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
15. The encoding method of claim 12 , further comprising generating a list of most probable modes used in intra prediction for a subsequent coding block to the current coding block according to the second prediction mode.
16. The encoding method of claim 15 , further comprising encoding the subsequent coding block using intra prediction with second information derived from prediction data determined based on at least one spatial neighboring block to the current coding block as a candidate.
17. The encoding method of claim 12 , wherein only when subsequent coding blocks and the current coding block belong to a same CTU, a same frame, or both, the subsequent coding blocks uses at least part of the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both.
18. The encoding method of claim 12 , wherein the stored first information derived from the first prediction mode, the second information referring to the second prediction mode, or both, is used for a subsequent coding block to reference.
19. The encoding method of claim 12 , further comprising:
combining the first prediction samples and the second prediction samples using a weighted sum to generate the single set of combined prediction samples.
20. The decoding method of claim 1 , wherein storing the first information derived from the first prediction mode, the second information referring to the second prediction mode, or both, comprises storing the first information derived from the first prediction mode.
21. The decoding method of claim 1 , wherein storing the first information derived from the first prediction mode, the second information referring to the second prediction mode, or both, comprises storing the second information derived from the second prediction mode.