Wavelet transform image coding using vector quantization

Abstract

Summary form only given. A novel scheme for image compression is proposed. Wavelet transform is used to obtain a set of orthonormal subclasses of images. Wavelets are functions that allow the construction of an orthonormal basis of L/sup 2/(R). The wavelet functions are well localized both in the space and frequency domains. The original image is decomposed on this orthonormal basis with a pyramidal algorithm architecture using quadrature mirror filters. This classification approach separates images (vectors) into perceptually distinct classes and thus matches the visual system model. The wavelet coefficients of each class are then vector quantized. The algorithm is based on a clustering approach and on the minimization of a distortion measure such as mean-squared error (MSE). A global codebook design unfortunately results in edge smoothing.<>