Motion compensation at a finer precision than motion vector differential
Extra precision can be added to motion vectors, such as those obtained using advanced motion vector prediction or frame rate upconversion, to produce improved motion compensation. The increased precision from neighboring blocks are used to provide increased precision to a motion vector for a current block. In various embodiments, the derivation of the increased precision can be found through various methods. In other embodiments, refined motion vectors can be saved for use in subsequent predictions, and a rate distorion optimized selection can be performed for determining whether or not to use the refinement.
1. A method, comprising:
obtaining an additional precision value for a motion vector, having an initial precision, for a block of video data from at least one neighboring block previously encoded;
refining the initial precision of the motion vector for the block of video data by assigning the additional precision value to the initial precision of the motion vector for the block of video data;
refining a plurality of motion vectors for a first row and/or column of sub-blocks of the block of video data using the refined motion vector for the block of video data and template matching;
refining a plurality of motion vectors for a second row and/or column of sub-blocks of the block of video data using an average of spatial neighboring motion vectors and the refined motion vector for the block of video data;
performing motion compensation for the block of video data by using the refined motion vectors of the sub-blocks; and
encoding said motion compensated block of video data.
2. The method of claim 1 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a lower fractional part of one motion vector, having a finer precision than the initial precision, from one neighboring block previously encoded that is a motion vector predictor for the block.
3. The method of claim 1 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a combination of lower fractional parts from at least two motion vectors from at least two neighboring blocks previously encoded.
4. The method of claim 1 , wherein the additional precision value is derived from a first N encountered motion vectors corresponding to neighboring blocks that have greater precision than the motion vector, having the initial precision, for the block of video data.
5. The method of claim 1 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N motion vectors corresponding to neighboring blocks that are closest in value to the motion vector, having the initial precision, for the block of video data.
6. A method, comprising:
obtaining an additional precision value for a motion vector, having an initial precision, for a block of video data from at least one neighboring block previously decoded;
refining the initial precision of the motion vector for the block of video data by assigning the additional precision value to the initial precision of the motion vector for the block of video data;
refining a plurality of motion vectors for a first row and/or column of sub-blocks of the block of video data using the refined motion vector for the block of video data and template matching;
refining a plurality of motion vectors for a second row and/or column of sub-blocks of the block of video data using an average of spatial neighboring motion vectors and the refined motion vector for the block of video data;
performing motion compensation for the block of video data by using the refined motion vectors of the sub-blocks; and
decoding said motion compensated block of video data.
7. The method of claim 6 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a lower fractional part of one motion vector, having a finer precision than the initial precision, from one neighboring block previously decoded that is a motion vector predictor for the block.
8. The method of claim 6 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a combination of lower fractional parts from at least two motion vectors from at least two neighboring blocks previously decoded.
9. The method of claim 6 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N encountered motion vectors corresponding to neighboring blocks that have greater precision than the motion vector, having the initial precision, for the block of video data.
10. The method of claim 6 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N motion vectors corresponding to neighboring blocks that are closest in value to the motion vector, having the initial precision, for the block of video data.
11. An apparatus, comprising:
a memory, and a processor configured to:
obtain an additional precision value for a motion vector, having an initial precision, for a block of video data from at least one neighboring block previously encoded,
refine the initial precision of the motion vector for the block of video data by assigning the additional precision value to the initial precision of the motion vector for the block of video data,
refine a plurality of motion vectors for a first row and/or column of sub-blocks of the block of video data using the refined motion vector for the block of video data and template matching,
refine a plurality of motion vectors for a second row and/or column of sub-blocks of the block of video data using an average of spatial neighboring motion vectors and the refined motion vector for the block of video data,
perform motion compensation for the block of video data by using the refined motion vectors of the sub-blocks, and
encode said motion compensated block of video data.
12. The apparatus of claim 11 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a lower fractional part of one motion vector, having a finer precision than the initial precision, from one neighboring block previously encoded that is a motion vector predictor for the block.
13. The apparatus of claim 11 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a combination of lower fractional parts from at least two motion vectors from at least two neighboring blocks previously encoded.
14. The apparatus of claim 11 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N encountered motion vectors corresponding to neighboring blocks that have greater precision than the motion vector, having the initial precision, for the block of video data.
15. The apparatus of claim 11 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N motion vectors corresponding to neighboring blocks that are closest in value to the motion vector, having the initial precision, for the block of video data.
16. An apparatus, comprising:
a memory, and a processor configured to:
obtain an additional precision value for a motion vector, having an initial precision, for a block of video data from at least one neighboring block previously decoded,
refine the initial precision of the motion vector for the block of video data by assigning the additional precision value to the initial precision of the motion vector for the block of video data,
refine a plurality of motion vectors for a first row and/or column of sub-blocks of the block of video data using the refined motion vector for the block of video data and template matching,
refine a plurality of motion vectors for a second row and/or column of sub-blocks of the block of video data using an average of spatial neighboring motion vectors and the refined motion vector for the block of video data,
perform motion compensation for the block of video data by using the refined motion vectors of the sub-blocks, and
decode said motion compensated block of video data.
17. The apparatus of claim 16 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a lower fractional part of one motion vector, having a finer precision than the initial precision, from one neighboring block previously decoded that is a motion vector predictor for the block.
18. The apparatus of claim 16 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is a combination of lower fractional parts from at least two motion vectors from at least two neighboring blocks previously decoded.
19. The apparatus of claim 16 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N encountered motion vectors corresponding to neighboring blocks that have greater precision than the motion vector, having the initial precision, for the block of video data.
20. The apparatus of claim 16 , wherein the additional precision value for the motion vector, having the initial precision, for the block of video data is derived from a first N motion vectors corresponding to neighboring blocks that are closest in value to the motion vector, having the initial precision, for the block of video data.