Block size dependent filter selection for motion compensation
View Patent ↗Disclosed are various embodiments for determining a filter length according to a width and height of the reference block and selecting a filter based on the block size. Corresponding pixels from memory may be fetched based on the filter length and the block size. Reference pixels may be interpolated based on the selected filter to generate a prediction.
1. A system, comprising:
a decoded picture buffer;
a pixel fetcher configured to:
determine, for each of a plurality of filters having different filter lengths, a memory expansion based on a width and a height of a reference block of a plurality of corresponding pixels from the decoded picture buffer;
select a filter of the plurality of filters based on the determined memory expansion for each of the plurality of filters; and
fetch the corresponding pixels from the decoded picture buffer based on the selected filter length and the block size; and
a multi-filter interpolator communicatively coupled to the fetcher, the multi-filter interpolator configured to:
select the filter of the plurality of filters based on the block size or the filter type selected by the pixel fetcher; and
interpolate reference pixels based on the selected filter to generate a prediction.
2. The system of claim 1 , wherein the system is embodied in an encoder or a decoder.
3. The system of claim 1 , wherein the block size dependent reference pixel fetcher is further configured to disable N×M motion blocks according to a block size restriction.
4. The system of claim 3 , wherein N and M are each equal to four (4).
5. The system of claim 3 , wherein N is equal to either four (4) or eight (8) and M is equal to either four (4) or eight (8).
6. The system of claim 1 , wherein the block size dependent reference pixel fetcher is further configured to disable bi-directional prediction for N×M motion blocks.
7. A method, comprising:
selecting, by circuitry of a pixel fetcher, a filter from a plurality of filters, each filter having a different length, according to a memory expansion associated with a block size of a reference block and the length of the filter;
fetching, by the circuitry of the pixel fetcher, corresponding pixels from memory based on the length of the selected filter and the block size; and
interpolating, by circuitry of a multi-filter interpolator, reference pixels based on the selected filter to generate a prediction.
8. The method of claim 7 , wherein the block size is based at least in part on a height and a width of the reference block.
9. The method of claim 7 , further comprising disabling N×M motion blocks according to a block size restriction.
10. The method of claim 7 , further comprising disabling bi-directional prediction for N×M motion blocks.
11. The method of claim 10 , wherein N is equal to either four (4) or eight (8) and M is equal to either four (4) or eight (8).
12. An apparatus, comprising:
a decoded picture buffer;
a pixel fetcher configured to:
determine, for each of a plurality of filters having different filter lengths, a memory expansion based on a width and a height of a reference block of a plurality of corresponding pixels from the decoded picture buffer;
select a filter of the plurality of filters based on the determined memory expansion for each of the plurality of filters; and
fetch the corresponding pixels from the decoded picture buffer based on the selected filter length and the block size; and
a multi-filter interpolator coupled to the fetcher, the multi-filter interpolator configured to:
select the filter of the plurality of filters based on the block size or the filter type from the fetcher; and
interpolate reference pixels based on the selected filter to generate a prediction.
13. The apparatus of claim 12 , wherein the block size dependent reference pixel fetcher is further configured to reject the block size according to a block size restriction.
14. The apparatus of claim 12 , further comprising disabling N×M motion blocks.
15. The apparatus of claim 12 , wherein the system is embodied in an encoder or a decoder.
16. The apparatus of claim 13 , wherein N and M are each equal to four (4).
17. The apparatus of claim 13 , wherein N is equal to either four (4) or eight (8) and M is equal to either four (4) or eight (8).
18. The apparatus of claim 14 , wherein N and M are each equal to four (4).
19. The system of claim 3 , wherein the pixel fetcher is further configured to select the width and height of the reference block from a plurality of predetermined widths and heights; and wherein disabling N×M motion blocks further comprises excluding the values N and M from the plurality of predetermined widths and heights.
20. The method of claim 9 , wherein disabling N×M motion blocks further comprises excluding the values N and M from a plurality of predetermined widths and heights, the block size of the reference block selected from the plurality of predetermined widths and heights.