Block-matching motion estimation technique for video compression of noisy source signals

摘要

A method for operating a data processing system to compress a first frame in an ordered sequence of image frames. A second frame precedes the first frame in the sequence of frames, and the second frame is preceded by a third frame in the sequence of frames. Each frame is divided into a plurality of processing blocks. The method first determines a displacement vector for at least one block in the second frame. The displacement vector identifies a region in the third frame that is an approximation of the block in the second frame. The displacement vectors in the second frame and measured vectors in the first frame are then used to determine an approximation for a displacement vector associated with one of the blocks in the first frame. This approximation is then used to define a search region in a frame preceding the first frame. The search region includes a plurality of sub-regions, each of the sub-regions having the same size as that block. The search region is then searched for the sub-region that best approximates the corresponding block in the first frame. In one embodiment of the present invention, the approximation to the displacement vector in the first frame is determined by forming a weighted sum of displacement vectors determined for the preceding frames. The coefficients of the weighted sums are determined by fitting the observed displacement vectors for a number of frames in the sequence.