Signal decomposition and coding using a multiresolution transform

Abstract

Signal representation and data coding for one and multidimensional signals have recently received considerable attention due to their importance to several modern technologies. Many useful contributions have been reported that employ wavelets and transform methods. Block transforms, particularly the discrete cosine transform, have been used in image-video coding. Signal decomposition has widely been used in conjunction with the discrete cosine transform for signal compression. In this paper, we explore the approximate trigonometric expansions for the purpose of signal decomposition and coding. Specifically, we give system interpretation to the approximate Fourier expansion using harmonic analysis. Furthermore, we apply the approximate trigonometric expansions to multispectral imagery, and investigate the potential of adaptive coding using blocks of images. The variable length basis functions computed by varying the user-defined parameter of the approximate trigonometric expansions are used for adaptive transform coding of images. Based on signal statistics, the proposed algorithm switches between a transform coder and a subband coder. It is shown that these expansions can be implemented by fast Fourier transform algorithm. Sample results for representing multidimensional signals are given to illustrate the efficiency of the proposed method. For comparison purposes, the results will be compared with techniques using block discrete cosine transform.