Method and apparatus for generalized OBMC
A block of video data that is split into sub-blocks forms a prediction for that sub-block using predictions from neighboring sub-blocks such that a current prediction for the sub-block is combined with weighted versions of the neighboring predictions. The neighboring sub-blocks motion vectors are checked to determine whether they are different than the motion vector of the sub-block being predicted. If so, than that corresponding neighboring sub-block is used to generate the prediction of the current sub-block. In an embodiment, two lines or two columns of pixels within a sub-block are used in forming the prediction for the current sub-block when the size of the block containing the sub-block is under a particular size.
1. A method, comprising:
determining whether motion vectors of horizontally and vertically neighboring sub-blocks of a current sub-block of a video coding block are different than a motion vector for the current sub-block, wherein the determining is based on checking the neighboring sub-blocks by pairs of neighboring sub-blocks and the pairs of neighboring sub-blocks include a first pair of horizontally neighboring sub-blocks left and right of the current sub-block and a second pair of vertically neighboring sub-blocks above and below the current sub-block;
filtering prediction pixels of said neighboring sub-blocks having different motion vectors than said current sub-block motion vector along with a current prediction of the current sub-block to generate a filtered prediction for the sub-block, wherein a number of prediction pixels to be filtered is set different for each direction and for each sub-block and can be set depending on size of each direction;
and if two opposite neighbors are available with different motion vector than a motion vector for a current sub-block, then two lines/columns are filtered from each neighbor and if one is available, then four lines/columns are filtered from the available neighbor; and
encoding said sub-block using said filtered prediction.
2. A non-transitory computer readable medium containing data content generated according to the method of claim 1 , for playback.
3. An apparatus for coding a block of video data, comprising:
a memory, and
a processor, configured to:
determine whether motion vectors of horizontally and vertically neighboring sub-blocks of a current sub-block of a video coding block are different than a motion vector for the current sub-block, wherein the determining is based on checking the neighboring sub-blocks by pairs of neighboring sub-blocks and the pairs of neighboring sub-blocks include a first pair of horizontally neighboring sub-blocks left and right of the current sub-block and a second pair of vertically neighboring sub-blocks above and below the current sub-block;
filter prediction pixels of said neighboring sub-blocks having different motion vectors than said current sub-block motion vector along with a current prediction of the current sub-block to generate a filtered prediction for the sub-block, wherein a number of prediction pixels to be filtered is set differently for each direction and for each sub-block and can be set depending on size of each direction;
and if two opposite neighbors are available with different motion vector than a motion vector for a current sub-block, then two lines/columns are filtered from each neighbor and if one is available, then four lines/columns are filtered from the available neighbor; and
encode said sub-block using said filtered prediction.
4. A method, comprising:
determining whether motion vectors of horizontally and vertically neighboring sub-blocks of a current sub-block of a video coding block are different than a motion vector for the current sub-block, wherein the determining is based on checking the neighboring sub-blocks by pairs of neighboring sub-blocks and the pairs of neighboring sub-blocks include a first pair of horizontally neighboring sub-blocks left and right of the current sub-block and a second pair of vertically neighboring sub-blocks above and below the current sub-block;
filtering prediction pixels of said neighboring sub-blocks having different motion vectors than said current sub-block motion vector along with a current prediction of the current sub-block to generate a filtered prediction for the sub-block, wherein a number of prediction pixels to be filtered is set differently for each direction and for each sub-block and can be set depending on size of each direction;
and if two opposite neighbors are available with different motion vector than a motion vector for a current sub-block, then two lines/columns are filtered from each neighbor and if one is available, then four lines/columns are filtered from the available neighbor; and
decoding said sub-block using said filtered prediction.
5. The method of claim 4 , wherein the filtering comprises using a prediction from a left or a top sub-block for two first pixels or two first lines, respectively, of the sub-block being predicted.
6. The method of claim 5 , wherein the two first pixels or two first lines of said sub-block prediction are filtered using a prediction from a left or a top neighboring sub-block, respectively, when a size of a block comprising said sub-block is less than a particular size.
7. The method of claim 4 , wherein the filtering comprises using a prediction from a right or a bottom sub-block for two last pixels or two last lines, respectively, of the sub-block being predicted.
8. The method of claim 4 , wherein the filtering comprises using one neighboring sub-block prediction when another one of said corresponding horizontal or vertical pairs of motion vectors is the same as the motion vector of said sub-block.
9. The method of claim 4 , wherein the filtering comprises using a prediction from a left or a top sub-block for a first pixel or first line, respectively, of the sub-block being predicted.
10. The method of claim 4 , wherein the filtering comprises using a prediction from a right or a bottom sub-block for a last pixel or last line, respectively, of the sub-block being predicted.
11. The method of claim 4 , wherein the filtering comprises two pixels or two lines of a sub-block being filtered for predictions corresponding to blocks comprising said sub-block larger than a particular size and one pixel or one line being filtered for predictions corresponding to blocks comprising said sub-block smaller than a particular size.
12. The method of claim 4 , wherein said current sub-block is part of a coding unit.
13. An apparatus for coding a block of video data, comprising:
a memory, and
a processor, configured to:
determine whether motion vectors of horizontally and vertically neighboring sub-blocks of a current sub-block of a video coding block are different than a motion vector for the current sub-block, wherein the determining is based on checking the neighboring sub-blocks by pairs of neighboring sub-blocks and the pairs of neighboring sub-blocks include a first pair of horizontally neighboring sub-blocks left and right of the current sub-block and a second pair of vertically neighboring sub-blocks above and below the current sub-block;
filter prediction pixels of said neighboring sub-blocks having different motion vectors than said current sub-block motion vector along with a current prediction of the current sub-block to generate a filtered prediction for the sub-block, wherein a number of prediction pixels to be filtered is set differently for each direction and for each sub-block and can be set depending on size of each direction;
and if two opposite neighbors are available with different motion vector than a motion vector for a current sub-block, then two lines/columns are filtered from each neighbor and if one is available, then four lines/columns are filtered from the available neighbor; and
decode said sub-block using said filtered prediction.
14. The apparatus of claim 13 , wherein the filtering comprises using a prediction from a left or a top sub-block for two first pixels or two first lines, respectively, of the sub-block being predicted.
15. The apparatus of claim 14 , wherein the two first pixels or two first lines of said sub-block prediction are filtered using a prediction from a left or a top neighboring sub-block, respectively, when a size of a block comprising said sub-block is less than a particular size.
16. The apparatus of claim 13 , wherein the filtering comprises using a prediction from a right or a bottom sub-block for two last pixels or two last lines, respectively, of the sub-block being predicted.
17. The apparatus of claim 13 , wherein the filtering comprises using one neighboring sub-block prediction when another one of said corresponding horizontal or vertical pairs of motion vectors is the same as the motion vector of said sub-block.
18. The apparatus of claim 13 , wherein the filtering comprises using a prediction from a left or a top sub-block for a first pixel or first line, respectively, of the sub-block being predicted.
19. The apparatus of claim 13 , wherein the filtering comprises using a prediction from a right or a bottom sub-block for a last pixel or last line, respectively, of the sub-block being predicted.
20. The apparatus of claim 13 , wherein the filtering comprises two pixels or two lines of a sub-block being filtered for predictions corresponding to blocks comprising said sub-block larger than a particular size and one pixel or one line being filtered for predictions corresponding to blocks comprising said sub-block smaller than a particular size.