INFORMATION-THEORY, WAVELETS, AND IMAGE COMPRESSION

We examine practical, theoretical, and speculative aspects oi wavelet transform-based image compression. Section I summarizes objectives and compares experimental results using a JPEG-standard cosine-based algo rithm with a wavelet based algorithm developed at ISS. Section II anal yzes image compression requirements using information theory to explai n why wavelet transform-based image compression works well. The wavele t transform is shown to be a simple transform that effectively exploit s second-order image statistics. Section III speculates about next-gen eration image compression and pattern recognition. It outlines a resea rch plan to develop a probabilistic image model that incorporates high er-order image statistics by using wavelet expansions to provide a con vergent series of finite dimensional marginal image probability densit ies. Physicists have successfully used similar cell cluster expansions to analyze lattice fields, Ising models, and Euclidean quantum fields . (C) 1996 John Wiley & Sons, Inc.