Space-variant regular decomposition quadtrees in adaptive interframe coding

Quadtrees are a compact hierarchical method of representation of images. The authors present a new type of scene adaptive coder which is based on a quadtree mean decomposition of the motion compensated frame-to-frame difference signal followed by a scalar quantization and variable wordlength encoding of the local sample means. As a fundamental property, the displacement vector is determined in a sense such that the resulting quadtree decomposition attains a minimum in terms of a minimum number of bits required for encoding. This underlying joint optimization of displacement and decomposition has proven excellent results for coding of color image sequences at a rate of 64 kbit/s. The new quadtree scene coder is superior over traditional transform-based concepts and attracts a regular algorithm structure which is well suited to VLSI realization