Application Specific Efficient VLSI Architectures for Orthogonal Single- and Multiwavelet Transforms

Abstract

In this paper, efficient VLSI architectures for orthogonal wavelet transforms with respect to common applications are presented. One class of orthogonal wavelet transforms is the singlewavelet transform which is based on one scaling and one wavelet function. An important application of this transform is signal denoising for which an efficient VLSI implementation and layout is derived in this paper. Contrary to singlewavelets, orthogonal multiwavelets are based on several scaling and wavelet functions. Since they allow properties like compact support, regularity, orthogonality and symmetry, simultaneously, being impossible in the singlewavelet case, multiwavelets are well suited bases for image compression applications. With respect to an efficient implementation of these orthogonal wavelet transforms, approximations of the exact rotation angles of the corresponding wavelet lattice filters are used. The approximations are realized by elementary CORDIC rotations. This method reduces the number of shift and add operations significantly with no influence on the good performance of the transforms. VLSI architectures for the computationally cheap transforms and related implementation aspects are discussed and design examples from architectural level down to layout are given.