Reference picture re-sampling
Predictive coding techniques may include resampling of reference pictures, where various coding parameters are determined based on the resolution(s) or pixel format(s) of the prediction references. In a first aspect, lists of weights for use in weighted prediction are based on the resolution(s) of prediction references. In a second aspect, resampling filter parameters are selected based on the resolutions of prediction references. In a third aspect, deblocking filter parameters are based on the resolution(s) of prediction references.
1. An image processing method, comprising:
decoding reference picture(s) having corresponding reference resolution(s);
determining prediction weight(s) for the reference picture(s) based on the corresponding reference resolutions(s);
resampling the decoded reference picture(s); and
predicting a portion of a current image based on the resampled reference picture(s) and the prediction weight(s).
2. The method of claim 1 , wherein:
the predicting is multihypothesis prediction with two reference pictures; and
the prediction weights depend on the difference in resolution of the two reference pictures.
3. The method of claim 2 , wherein,
a first prediction weight is determined for a first reference picture having a first resolution; and
a second prediction weight, lower than the first prediction weight, is determined for a second reference picture having a second resolution lower than the first resolution.
4. The method of claim 1 , wherein, when at least one of the corresponding reference resolutions is smaller than a current resolution of the current image, then all the prediction weights are determined to be the same weight.
5. The method of claim 1 , wherein the prediction weights for the reference picture(s) are determined by selecting an entry from a list of prediction weights, and wherein the prediction weights are encoded in a bitstream along with an indication of the relative probabilities of the weights in the list of prediction weights.
6. The method of claim 1 , wherein the prediction weights for the reference picture(s) are determined by selecting an entry from a list of prediction weights based on the corresponding reference resolution(s).
7. The method of claim 1 , wherein the prediction weights for the reference picture(s) are determined by selecting an entry from a list of prediction weights, and the list of prediction weights is selected from a plurality of lists of weights based on the corresponding reference resolution(s).
8. The method of claim 1 , wherein
the prediction reference(s) are classified as based on whether their corresponding reference resolutions(s) are larger or smaller than a current resolution of the current image; and
the prediction weights for the reference picture(s) are determined based on the classification(s).
9. A system for image processing, comprising:
a decoder for decoding reference pictures having corresponding reference resolution(s);
a reference picture buffer storing the decoded reference pictures;
a resampler for resampling reference pictures from the reference picture buffer;
a controller for prediction weight(s) for the reference picture(s) based on the corresponding reference resolution(s); and
a predictor for predicting a current image based on the resampled reference pictures and the prediction weights.
10. The system of claim 9 , wherein:
the predicting is multihypothesis prediction with two reference pictures;
the prediction weights depend on the difference in resolution of the two reference pictures;
a first prediction weight is determined for a first reference picture having a first resolution; and
a second prediction weight, lower than the first prediction weight, is determined for a second reference picture having a second resolution lower than the first resolution.
11. A non-transitory computer readable medium having instructions that, when executed by a processor, cause:
decoding reference picture(s) having corresponding reference resolution(s);
determining prediction weight(s) for the reference picture(s) based on the corresponding reference resolution(s);
resampling the decoded reference picture(s); and
predicting a portion of a current image based on the resampled reference picture(s) and the prediction weight(s).
12. The medium of claim 11 , wherein:
the predicting is multihypothesis prediction with two reference pictures;
the prediction weights depend on the difference in resolution of the two reference pictures;
a first prediction weight is determined for a first reference picture having a first resolution; and
a second prediction weight, lower than the first prediction weight, is determined for a second reference picture having a second resolution lower than the first resolution.