PREDICTIVE CLASSIFIED VECTOR QUANTIZATION FOR IMAGE CODING.

Abstract

A novel method of vector quantization, called predictive classified vector quantization (PCVQ), is proposed for image coding. It utilizes spatial correlations across the boundaries of pixel blocks to achieve a high compression rate. The edge/shade and mean intensity information of each block can be predicted and classified. Each block is coded by a quantizer belonging to the same class. The PCVQ consists of compact subcodebooks and gives satisfactory performance at 0. 375 b/pixel.