No label defined (Q4218367)

We present an overview of Coifman wavelet systems, a family of wavelet systems with very nice properties both in the theoretical sense and application sense. The original idea goes back to R. Coifman of Yale University. In the spring of 1989, he suggested that it might be worthwhile to construct orthogonal wavelet systems with vanishing moments not only for the wavelet functions (which is the hypothesis posed on the Daubechies wavelet systems), but also for the scaling functions. This turned out to be an interesting success. One key property of these orthogonal wavelet systems with vanishing moments equally distributed between scaling functions and wavelet functions (which we call orthogonal Coifman wavelet systems) is that they have very good approximation properties with exponential decay. Biorthogonal Coifman wavelet systems, the biorthogonal counterparts of orthogonal Coifman wavelet systems, also have good approximation properties with exponential decay. An attractive feature of biorthogonal Coifman wavelet systems is that all the scaling filters have dyadic rational coefficients. Thus, we have obtained a family of biorthogonal wavelet systems with a multiplication-free discrete wavelet transform. Moreover, they are compactly supported, and converge to the Sinc wavelet system. Half of the members (more precisely, those with odd degrees) of the family are symmetric, which indicates the potential promising of these biorthogonal Coifman wavelet systems in the applications of signal processing. Actually, based on a complete coding performance evaluation, biorthogonal Coifman wavelet systems are very useful for image transform coding and seem to be quite comparable to the wavelet systems used in the state-of-the-art compression systems. For the entire collection see [Zbl 0889.00026].