BEWEGUNGSSCHÄTZUNGSSYSTEM UND -VERFAHREN

摘要

Techniques for doing motion estimation in video encoding. A predictor block which is a best match for a current block is located by comparing the current block of the picture currently being encoded with candidate blocks that are located using the motion vectors of blocks that are either spatially or temporally proximate to the current block. If a given proximate block does not have a usable motion vector, another proximate block is used. The temporally proximate blocks are blocks in the last picture to be encoded prior to the current picture. The candidate blocks are further used to define a minimized portion of the search space which will fit inside the local RAM of the processor used to compare the current block with the candidates. The minimized search space is used in a two-phase method which begins with the candidates located using the proximate blocks and continues with a diamond-shaped search space centered on the best candidate located using the proximate blocks. The duration of the search method is controlled by three thresholds. Each succeeding one of the thresholds requires a closer match for the predictor than the previous threshold. The size and shape of the diamond-shaped search space may vary according to the motion vectors of previous predictors and according to the number of processor cycles that are available to perform the search. Similarly, the number and locations of the proximate blocks may vary according to the motion vectors of previous predictors and the number of processor cycles available.