Reusing reference sample memory based on flexible partition type
A method for video decoding includes determining, based on prediction information, whether the current block uses a flexible partition type. Reconstructed samples of the reference block are stored in a cache memory, which stores reconstructed samples of coding regions of a first size comprising a coding tree unit (CTU) of the current block and reconstructed samples of at least one coding region of the first size comprising a CTU adjacent to the CTU of the current block. The cache memory stores either (i) only reconstructed samples of the coding regions of the CTU of the current block or (ii) reconstructed samples of coding regions of a second size of the CTU of the current block and reconstructed samples of at least one coding region of the second size of the CTU adjacent to the CTU of the current block.
1. A method for video encoding, comprising:
determining a block vector that points to a reference block in a same picture as a current block, reconstructed samples of the reference block being stored in a cache memory, which stores reconstructed samples of a first coding tree unit (CTU) in which the current block is located and reconstructed samples of a second CTU adjacent to the first CTU;
encoding samples of the current block based on the reconstructed samples of the reference block stored in the cache memory; and
updating the cache memory by storing reconstructed samples of the current block in the cache memory such that the reconstructed samples of the second CTU are not available for reconstruction of a next block,
wherein a search range of reconstructed samples of the second CTU stored in the cache memory and available for encoding a respective block of the first CTU is different for each block of the first CTU and dependent on a location of the respective block with respect to reconstructed blocks within the first CTU.
2. The method of claim 1 , wherein each block has a square shape.
3. The method of claim 1 , wherein each CTU is configured to have 128 samples by 128 samples of a luma component.
4. The method of claim 1 , wherein the cache memory is updated after all of the reconstructed samples of the current block are available, by storing all of the reconstructed samples of the current block in the cache memory and by removing from the cache memory all reconstructed samples of one coding region comprising the second CTU adjacent to the first CTU.
5. The method of claim 1 , further comprising:
checking whether multiple corners of the reference block are within a coding region.
6. The method of claim 5 , wherein
each CTU contains 128×128 luma samples and each block is 64×64 luma samples in size.
7. The method of claim 5 , wherein
each CTU contains less than 128×128 luma samples and each block is the same as a CTU in size.
8. The method of claim 1 , further comprising:
when the reconstructed samples of the reference block are partially available in the cache memory, updating the cache memory such that all reconstructed samples of the reference block are available in the cache memory.
9. The method of claim 8 , further comprising:
when the reconstructed samples of the reference block are partially available in the cache memory, moving the reference block in a direction to reduce a distance between the reference block and the current block.
10. The method of claim 8 , further comprising;
moving the reference block in a direction that increases a percentage of reference samples in a coding region.
11. The method of claim 1 , further comprising:
using information of a first reference sample that is in the cache memory as information of a second reference sample that is not available in the cache memory when the reference block is partially available in the cache memory.
12. The method of claim 1 , wherein flexible block partition is prohibited as a partition of each CTU.
13. An apparatus of video decoding, comprising:
processing circuitry configured to:
determine a block vector that points to a reference block in a same picture as a current block, reconstructed samples of the reference block being stored in a cache memory, which stores reconstructed samples of a first coding tree unit (CTU) in which the current block is located and reconstructed samples of a second CTU adjacent to the first CTU;
reconstruct samples of the current block based on the reconstructed samples of the reference block stored in the cache memory; and
update the cache memory by storing reconstructed samples of the current block in the cache memory such that the reconstructed samples of the second CTU are not available for reconstruction of a next block,
wherein a search range of reconstructed samples of the second CTU stored in the cache memory and available for reconstructing a respective block of the first CTU is different for each block of the first CTU and dependent on a location of the respective block with respect to reconstructed blocks within the first CTU.
14. The apparatus of claim 13 , wherein each block has a square shape.
15. The apparatus of claim 13 , wherein each CTU is configured to have 128 samples by 128 samples of a luma component.
16. The apparatus of claim 13 , wherein the cache memory is updated after all of the reconstructed samples of the current block are available, by storing all of the reconstructed samples of the current block in the cache memory and by removing from the cache memory all reconstructed samples of one coding region comprising the second CTU adjacent to the first CTU.
17. The apparatus of claim 13 , wherein the processing circuitry is further configured to:
check whether multiple corners of the reference block are within a coding region.
18. The apparatus of claim 17 , wherein
each CTU contains 128×128 luma samples and each block is 64×64 luma samples in size.
19. The apparatus of claim 17 , wherein
each CTU contains less than 128×128 luma samples and each block is the same as a CTU in size.
20. A method of processing visual media data, the method comprising:
processing a bitstream of the visual media data according to a format rule,
wherein the bitstream indicates a block vector that points to a reference block in a same picture as a current block,
wherein the format rule specifies that reconstructed samples of the reference block are stored in a cache memory, which stores reconstructed samples of a first coding tree unit (CTU) in which the current block is located and reconstructed samples of a second CTU adjacent to the first CTU,
wherein the format rule specifies that the cache memory is updated by storing reconstructed samples of the current block in the cache memory such that the reconstructed samples of the second CTU are not available for reconstruction of a next block, and
wherein the format rule specifies that a search range of reconstructed samples of the second CTU stored in the cache memory and available for encoding a respective block of the first CTU is different for each block of the first CTU and dependent on a location of the respective block with respect to reconstructed blocks within the first CTU.