Video decoder and a video encoder using motion-compensated prediction

摘要

A video decoder using motion-compensated prediction has a predictor configured to predict a first subset of parameters of a plurality of motion parameters, defining the motion-compensated prediction of a predetermined region of a picture of a video, to obtain a prediction for the first subset. The video decoder further has a reconstructor configured to reconstruct the first subset of parameters from the prediction of the first subset and residual information for the first subset contained in a data stream, to obtain a reconstruction of the first subset. The predictor is further configured to predict a second subset of parameters of the plurality of motion parameters, being disjoint from the first subset, using the reconstruction of the first subset.

主权项

1. A video decoder using motion-compensated prediction comprising:
a predictor configured to predict a first subset of parameters of a plurality of motion parameters defining the motion-compensated prediction of a predetermined region of a picture of a video to acquire a prediction for the first subset; and a reconstructor configured to reconstruct the first subset of parameters from the prediction for the first subset and residual information for the first subset comprised in a data stream, to acquire a reconstruction of the first subset, wherein the predictor is configured to predict a second subset of parameters of the plurality of motion parameters, being disjoint from the first subset using the reconstruction of the first subset, to acquire a prediction for the second subset; the predictor is configured to determine a first candidate set of previously decoded motion information, each motion information defining a plurality of motion parameters defining the motion-compensated prediction for a region neighboring the predetermined region or defining a motion-compensated prediction for the predetermined region in a previously decoded picture, to acquire the prediction for the first subset; the predictor is configured to select a second candidate set as a subset of the first candidate set of previously decoded motion information based on the reconstruction of the first subset, and to acquire the prediction for the second subset using the second candidate set; and at least one of the predictor and the reconstructor comprises a hard implementation.