A self-organizing adaptive vector quantization technique

Abstract

This paper describes a new adaptive vector quantization scheme suitable for encoding monochrome and color images. In the proposed technique, a large codebook, consisting of two sections, called higher and lower priority sections, respectively, and representing common and specific characteristics of images, is designed. During encoding the entries in the two sections are reorganized and exchanged as a function of frequency of codevector usage at given intervals. The rate and the extent of adaptation are dictated by the update interval and the desired level of quality, respectively. Adaptation is achieved without requiring any transmission of the vectors themselves. The codebook is expected to adaptively self-organize itself in such a manner that the highpriority part of the codebook is sensitized to track the local (specific) characteristics of the input image. This method improves the coding efficiency and provides a perceptually more consistent image quality throughout all areas of the restructured image. Simulation results show that for the test color images bit rates in the range of 0.15–0.25 bits per pixel with a signal-to-noise ratio around 27–30 dB can be achieved.