Real-Time Image Compression Based on Wavelet Vector Quantization, Algorithm and VLSI Architecture

Abstract

We present an algorithm and VLSI architecture of a real-time image compression that is based on vector quantization (VQ) of discrete wavelet transform (DWT) coefficients. Containing the hardware of both wavelet and VQ, this architecture is capable of performing real-time image compression for grayscale and 24 bit color images with a resolution of 1024/spl times/1024 pixels. The algorithm for 2D DWT is a lifting-based non-separable 2D DWT. Difference-codebook binary tree search VQ (DC-BTSVQ) is used in the algorithm which leads to a high speed encoder. Binary tree search VQ (BTSVQ) also requires low hardware overhead in VLSI implementation. The proposed real-time wavelet vector quantization (WVQ) algorithm combines multistage VQ (MSVQ) and classified VQ (CVQ). MSVQ encodes the lowest frequency subbands of the wavelet coefficients and has the advantage of robustness to wireless channel fading. CVQ encodes the higher frequency subbands and results in higher coding efficiency.