Chrominance vector quantization for cellular video-telephony

Abstract

The paper describes a vector quantization technique which can be added to a standard codec, like H.263, in order to improve its performance for very high compression. The main idea is to convert two chrominance components of a video-telephone signal into one scalar signal. This scalar chrominance signal is then compressed in a motion compensated coder. The crucial problem is proper ordering of the pixel numbers which are labels of the codebook entries. This ordering dramatically influences the performance of the coder.