Motion vector scaling for non-uniform motion vector grid

1. A method, comprising:

determining a motion vector for a second block that is a spatially neighboring block or a temporal co-located block to a first block, wherein values associated with the motion vector is are represented on a non-uniform motion vector grid having a first number of positions, the non-uniform motion vector grid having non-uniform sub-pixel phase offsets between integer pixels;

mapping, by a computing device, the values to a uniform motion vector grid, the uniform motion vector grid having a second number of positions greater than the first number of positions due to the presence of more sub-pixel positions between the integer pixels than the non-uniform motion vector grid and providing a uniform distribution of sub-pixel positions between the integer pixels;

scaling, by the computing device, the values represented on the uniform motion vector grid to generate scaled motion vector values; and

mapping, by the computing device, the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid, wherein the scaled motion vector values represented on the non-uniform motion vector grid are associated with the first block for a temporal prediction process.

2. The method of claim 1 , wherein the values represented on the uniform motion vector grid comprises scaling the values based on a difference between a first reference picture for the first block and a second reference picture for the second block.

3. The method of claim 2 , wherein the difference is a temporal difference.

4. The method of claim 1 , wherein mapping the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid comprises using an algorithm to perform the mapping to the non-uniform motion vector grid.

5. The method of claim 4 , wherein the algorithm includes mapping a subset of positions on the uniform motion vector grid to a single position on the non-uniform motion vector grid.

6. The method of claim 1 , further comprising:

coding information for the scaled motion vector values in a bitstream for the video content; and

sending the bitstream from an encoder to a decoder.

7. The method of claim 1 , wherein each component of the motion vector is represented on the non-uniform motion vector grid such that each motion vector component is located at a sub-pixel position determined by a remainder resulting from dividing the motion vector component by the first number of positions.

8. A method, comprising:

receiving a bitstream from an encoder at a decoder;

determining a motion vector for a second block that is a spatially neighboring block or a temporal co-located block to a first block using information in the bitstream, wherein values associated with the motion vector are represented on a non-uniform motion vector grid having a first number of positions, the non-uniform motion vector grid having non-uniform sub-pixel phase offsets between integer pixels;

mapping, by the decoder, the values to a uniform motion vector grid, the uniform motion vector grid having a second number of positions greater than the first number of positions due to the presence of more sub-pixel positions between the integer pixels than the non-uniform motion vector grid and providing a uniform distribution of sub-pixel positions between the integer pixels;

scaling, by the decoder, the values represented on the uniform motion vector grid to generate scaled motion vector values; and

mapping, by the decoder, the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid, wherein the scaled motion vector values represented on the non-uniform motion vector grid are associated with the first block for a temporal prediction process to decode the bitstream.

9. The method of claim 8 , wherein scaling the values represented on the uniform motion vector grid comprises scaling the values based on a temporal difference between a first reference picture for the first block and a second reference picture for the second block.

10. An apparatus, comprising:

one or more computer processors; and

a non-transitory computer-readable storage medium comprising instructions, that when executed, control the one or more computer processors to be configured for:

determining a motion vector for a second block that is a spatially neighboring block or a temporal co-located block to a first block, wherein values associated with the motion vector are represented on a non-uniform motion vector grid having a first number of positions, the non-uniform motion vector grid having non-uniform sub-pixel phase offsets between integer pixels;

mapping the values to a uniform motion vector grid, the uniform motion vector grid having a second number of positions greater than the first number of positions due to the presence of more sub-pixel positions between the integer pixels than the non-uniform motion vector grid and providing a uniform distribution of sub-pixel positions between the integer pixels;

scaling the values represented on the uniform motion vector grid to generate scaled motion vector values; and

mapping the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid, wherein the scaled motion vector values represented on the non-uniform motion vector grid are associated with the first block for a temporal prediction process.

11. The apparatus of claim 10 , wherein scaling the values represented on the uniform motion vector grid comprises scaling the values based on a difference between a first reference picture for the first block and a second reference picture for the second block.

12. The apparatus of claim 11 , wherein the difference is a temporal difference.

13. The apparatus of claim 10 , wherein mapping the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid comprises using an algorithm to perform the mapping to the non-uniform motion vector grid.

14. The apparatus of claim 13 , wherein the algorithm includes mapping a subset of positions on the uniform motion vector grid to a single position on the non-uniform motion vector grid.

15. The apparatus of claim 10 , further operable for:

coding information for the scaled motion vector values in a bitstream for the video content; and

sending the bitstream from an encoder to a decoder.

16. The apparatus of claim 10 , further operable for:

receiving a bitstream from an encoder at a decoder;

determining the motion vector for the second block using information in the bitstream; and

performing the scaling to determine the scaled motion vector values for a decoding process of the bitstream.

17. An apparatus, comprising:

one or more computer processors; and

a non-transitory computer-readable storage medium comprising instructions, that when executed, control the one or more computer processors to be configured for:

receiving a bitstream from an encoder at a decoder;

determining a motion vector for a second block that is a spatially neighboring block or a temporal co-located block to a first block using information in the bitstream, wherein values associated with the motion vector are represented on a non-uniform motion vector grid having a first number of positions, the non-uniform motion vector grid having non-uniform sub-pixel phase offsets between integer pixels;

mapping the values to a uniform motion vector grid, the uniform motion vector grid having a second number of positions greater than the first number of positions due to the presence of more sub-pixel positions between the integer pixels than the non-uniform motion vector grid and providing a uniform distribution of sub-pixel positions between the integer pixels;

scaling the values represented on the uniform motion vector grid to generate scaled motion vector values; and

mapping the scaled motion vector values represented on the uniform motion vector grid to the non-uniform motion vector grid, wherein the scaled motion vector values represented on the non-uniform motion vector grid are associated with the first block for a temporal prediction process to decode the bitstream.