초록 ▼

Taught herein are techniques to incorporate temporal motion vector prediction in encoding and decoding a video stream. Motion vectors are temporally predicted for blocks of a frame. An extrapolated position for at least one block of a first frame is predicted in a second frame using a respective mot

Taught herein are techniques to incorporate temporal motion vector prediction in encoding and decoding a video stream. Motion vectors are temporally predicted for blocks of a frame. An extrapolated position for at least one block of a first frame is predicted in a second frame using a respective motion vector of the at least one block. A temporal motion vector for a block of the second frame is calculated as a combination of respective motion vectors of blocks overlapping the block of the second frame if at least one of the overlapping blocks is an extrapolated block at the extrapolated position. The temporal motion vector is used in a motion estimation algorithm to select a motion vector for the block, and the block is encoded using the selected motion vector.

대표청구항 ▼

1. A method for encoding a video stream including a plurality of frames, the method comprising: predicting an extrapolated position for a plurality of blocks of a first frame in a second frame using a respective motion vector used for inter-prediction of each of the plurality of blocks of the first

1. A method for encoding a video stream including a plurality of frames, the method comprising: predicting an extrapolated position for a plurality of blocks of a first frame in a second frame using a respective motion vector used for inter-prediction of each of the plurality of blocks of the first frame, the second frame temporally subsequent to the first frame in a sequence of the plurality of frames;calculating a temporal motion vector for a first block of the second frame as a combination of respective motion vectors of blocks overlapping the first block of the second frame using a processor, wherein at least a first block of the overlapping blocks is an extrapolated block at the extrapolated position and at least a second block of the overlapping blocks includes an area of the block of the second frame not overlapped by any extrapolated blocks, the motion vector for the at least the second block of the overlapping blocks being a zero vector when calculating the temporal motion vector;using the temporal motion vector in a motion estimation algorithm to select a motion vector for the first block of the second frame; andencoding the first block of the second frame using the selected motion vector. 2. The method of claim 1, further comprising: setting a temporal motion vector for a second block of the second frame to zero when no extrapolated block overlaps the second block of the second frame. 3. The method of claim 1, further comprising: applying a filter to the respective motion vectors of the overlapping blocks before using the temporal motion vector in the motion estimation algorithm. 4. The method of claim 1, further comprising: applying a filter to the respective motion vectors of the overlapping blocks after calculating the temporal motion vector; andrecalculating the temporal motion vector before using the temporal motion vector in the motion estimation algorithm when at least one of the respective motion vectors is removed by the filter. 5. The method of claim 4 wherein the filter compares each of the respective motion vectors to the temporal motion vector. 6. The method of claim 4 wherein applying the filter comprises removing a motion vector mvk of a block k of the overlapping blocks if the difference between the motion vector mvk and the temporal motion vector is greater than a standard deviation of the motion vectors of the overlapping blocks. 7. The method of claim 1 wherein calculating the temporal motion vector comprises weighting the respective motion vectors of the overlapping blocks based on respective area of overlap between a respective overlapping block and the first block of the second frame. 8. The method of claim 1 wherein predicting the extrapolated position comprises predicting an extrapolated position for all blocks of the first frame except those blocks of the first frame referring to a reference frame other than a frame adjacent to the first frame in the sequence of the plurality of frames. 9. The method of claim 8, further comprising: reviewing each block of the second frame for an existence of overlapping extrapolated blocks; andcalculating a temporal motion vector for each block of the second frame wherein the temporal motion vector for the block of the second frame is a zero vector when no extrapolated block overlaps the block of the second frame. 10. The method of claim 1, further comprising: transmitting the encoded first block of the second frame and a code indicating which motion vector to select for decoding the first block of the second frame. 11. An apparatus for encoding a video stream having a plurality of frames, the apparatus comprising: a memory; anda processor configured to execute instructions stored in the memory to: predict an extrapolated position for each block of a plurality of blocks of a first frame in a second frame using a respective motion vector used for inter-prediction of each of the plurality of blocks of the first frame, the second frame temporally subsequent to the first frame in a sequence of the plurality of frames;calculate a temporal motion vector for a first block of the second frame as a combination of respective motion vectors of blocks overlapping the first block of the second frame, wherein at least a first block of the overlapping blocks is an extrapolated block at the extrapolated position and at least a second block of the overlapping blocks includes an area of the block of the second frame not overlapped by any extrapolated blocks, the motion vector for the at least the second block of the overlapping blocks being a zero vector when calculating the temporal motion vector;use the temporal motion vector in a motion estimation algorithm to select a motion vector for the first block of the second frame; andencode the first block of the second frame using the selected motion vector. 12. The apparatus of claim 11 wherein the processor further comprises: a filter that receives the respective motion vectors of the overlapping blocks and the temporal motion vector, compares each of the respective motion vectors to the temporal motion vector and removes a motion vector of the respective motion vectors from a re-calculation of the temporal motion vector when a difference between that motion vector and the temporal motion vector is higher than a desired difference. 13. The apparatus of claim 11 wherein the processor is configured to set a temporal motion vector for a second block of the second frame to a zero vector when no extrapolated blocks overlap the second block of the second frame. 14. The apparatus of claim 11 wherein the processor is configured to use the temporal motion vector in a motion estimation algorithm by: selecting one of a plurality of coding modes to encode the first block of the second frame, each of the plurality of coding modes indicating a respective motion vector to be used to encode the first block of the second frame, one of the respective motion vectors being the temporal motion vector for the first block of the second frame. 15. The apparatus of claim 11 wherein the processor is configured to use the temporal motion vector in a motion estimation algorithm by: finding a best motion vector correlating a prediction block of the first frame with the first block of the second frame; andcalculating a residual vector for the first block of the second frame by subtracting the temporal motion vector from the best motion vector; and wherein the processor is configured to encode the residual vector. 16. The apparatus of claim 11 wherein the processor is configured to predict the extrapolated position (r,c)i of a block j of the first frame i−1 in the second frame i according to: (r,c)i=(r,c)i-1−mvi−1,j;wherein(r,c)i-1, is the position of the block j in the first frame i−1; andmvi-1,j is the motion vector of the block j in the first frame i−1. 17. A method for decoding a video stream including a plurality of frames, the method comprising: receiving a compressed bitstream including data representing a first frame, a second frame, motion vectors of the first frame and a code indicating how to select a motion vector for a block of the second frame;decoding a plurality of blocks of the first frame;predicting an extrapolated position for at least some of the plurality of decoded block of the first frame in the second frame using a respective motion vector used for inter-prediction of each of the at least some decoded blocks;calculating a temporal motion vector for a first block of the second frame as a combination of respective motion vectors of blocks overlapping the first block of the second frame using a processor, wherein at least a first block of the overlapping blocks is an extrapolated block at the extrapolated position and at least a second block of the overlapping blocks includes an area of the block of the second frame not overlapped by any extrapolated blocks, the motion vector for the at least the second block of the overlapping blocks being a zero vector when calculating the temporal motion vector;using the temporal motion vector in a motion estimation algorithm to select a motion vector for the first block of the second frame according to the code associated with the first block of the second frame; anddecoding the first block of the second frame using the selected motion vector. 18. The method of claim 17, further comprising: applying a filter to the respective motion vectors of the overlapping blocks after calculating the temporal motion vector; andrecalculating the temporal motion vector before using the temporal motion vector in the motion estimation algorithm when at least one of the respective motion vectors is removed by the filter. 19. The method of claim 17 wherein using the temporal motion vector in a motion estimation algorithm comprises: using the code to select one of a plurality of coding modes to decode the first block of the second frame, the plurality of coding modes indicating a respective motion vector to be used to decode the first block of the second frame, one of the respective motion vectors being the temporal motion vector for the first block of the second frame.