A distortion analysis of image VQ-based coding using a finite mixture distribution model

Traditional coding schemes break an image into blocks prior to coding. It is possible to classify current algorithms according to the block construction. Either the block size is kept constant, or the block size is image dependent. Since most natural images can be divided into regions of high and low detail, variable block-size coding techniques exploit more efficiently the structure of the data. The superior performance of this scheme over fixed block ones has been observed experimentally. Rate-distortion analysis is carried out for compression systems using vector quantization. Using an appropriate model of the signal we derive analytical results that assess the superiority of variable block vector quantization algorithms.