Vector quantization video encoder using hierarchical cache memory scheme

Abstract

A system compresses image blocks via successive hierarchical stages and motion encoders which employ caches updated by stack replacement algorithms. Initially, a background detector compares the present image block with a corresponding previously encoded image block and if similar, the background detector terminates the encoding procedure by setting a flag bit. Otherwise, the image block is decomposed into smaller present image subblocks. The smaller present image subblocks are each compared with a corresponding previously encoded image subblock of comparable size within the present image block. When a present image subblock is similar to a corresponding previously encoded image subblock, then the procedure is terminated by setting a flag bit. Alternatively, the present image subblock is forwarded to a motion encoder where it is compared with displaced image subblocks, which are formed by displacing previously encoded image subblocks by motion vectors that are stored in a cache, to derive a first distortion vector. When the first distortion vector is below a first threshold T.sub.M, the procedure is terminated and the present image subblock is encoded by setting flag bit and a cache index corresponding to the first distortion vector. Alternatively, the present image subblock is passed to a block matching encoder where it is compared with other previously encoded image subblocks to derive a second distortion vector. When the second distortion vector is below a second threshold T.sub.m, the procedure is terminated by setting a flag bit, by generating the second distortion vector, and by updating the cache.