Prediction systems and methods for video coding based on filtering nearest neighboring pixels
Prediction systems and methods for video coding are described based on nearest neighboring pixels. In exemplary embodiments, to code a first pixel, a plurality of neighboring pixels of the first pixel are reconstructed. The coefficients of a filter such as a Wiener filter are derived based on the reconstructed neighboring pixels. The Wiener filter is applied to the reconstructed neighboring pixels to predict the first pixel. The coefficients of the Wiener filter may be derived on a pixel-by-pixel or a block-by-block basis. The reconstructed pixels may be pixels in the same picture (for intra prediction) or in a reference picture (for inter prediction). In some embodiments, the residuals of the prediction are encoded using RDPCM. In some embodiments, the residuals may be predicted using a Wiener filter.
1. A method comprising:
generating an intra prediction for each pixel in a current block of pixels;
determining coefficients of a filter associated with the current block of pixels;
applying the filter to a plurality of pixels, including at least one of the intra-predicted pixels in the current block, to yield a filter-based prediction for at least one of the pixels in the current block; and
reconstructing the pixels in the current block by adding a prediction residual to the filter-based prediction of at least one of the pixels in the current block.
2. The method of claim 1 , wherein the filter is a Wiener filter.
3. The method of claim 1 , wherein applying of the filter is performed only if a flag is signaled in a bitstream for the current block indicating that filter-based prediction is enabled.
4. The method of claim 1 , wherein determining coefficients of the filter comprises deriving the coefficients from neighboring reconstructed pixels.
5. The method of claim 1 , wherein determining coefficients of the filter comprises selecting filter coefficients to substantially minimize a mean square error between (i) reconstructed pixels in a template region neighboring the current block and (ii) a filtered version of an intra prediction of the pixels in the template region.
6. The method of claim 1 , wherein the coefficients are signaled explicitly in a bitstream.
7. The method of claim 1 , wherein determining the coefficients comprises selecting the coefficients from a lookup table.
8. The method of claim 7 , wherein the coefficients are selected from the lookup table based on a block size and on the intra prediction mode used to generate the intra prediction.
9. The method of claim 1 , further comprising performing in-loop filtering on the reconstructed pixels.
10. An apparatus comprising:
a processor configured to perform at least:
generating an intra or inter prediction for each pixel in a current block of pixels;
determining coefficients of a filter associated with the current block of pixels;
applying the filter to a plurality of pixels, including at least one of the intra-predicted or inter-predicted pixels in the current block, to yield a filter-based prediction for at least one of the pixels in the current block; and
reconstructing the pixels in the current block by adding a prediction residual to the filter-based prediction of at least one of the pixels in the current block.
11. The apparatus of claim 10 , wherein the processor is further configured to perform in-loop filtering on the reconstructed pixels.
12. The apparatus of claim 10 , wherein generating an intra or inter prediction comprises generating an inter prediction.
13. A method comprising:
generating a motion-compensated prediction for each pixel in a current block of pixels;
determining coefficients of a filter associated with the current block of pixels;
applying the filter to a plurality of pixels, including at least one of the motion-compensated predicted pixels in the current block to yield a filter-based prediction for at least one of the pixels in the current block; and
reconstructing the pixels in the current block by adding a prediction residual to the filter-based prediction of at least one of the pixels in the current block.
14. The method of claim 13 , wherein the filter is a Wiener filter.
15. The method of claim 13 , wherein determining coefficients of the filter comprises deriving the coefficients from neighboring reconstructed pixels.
16. The method of claim 13 , wherein determining coefficients of the filter comprises selecting filter coefficients to substantially minimize a mean square error between (i) reconstructed pixels in a template region neighboring the current block and (ii) a filtered version of a motion-compensated prediction of the pixels in the template region.
17. The method of claim 13 , wherein the coefficients are signaled explicitly in a bitstream.
18. The method of claim 13 , wherein determining the coefficients comprises selecting the coefficients from a lookup table.
19. The method of claim 13 , wherein the coefficients are selected from the lookup table based on a block size and on an edge direction of the motion-compensated prediction.
20. The method of claim 13 , further comprising performing in-loop filtering on the reconstructed pixels.