Iterative reweighted least squares and the design of two-dimensional FIR digital filters

Abstract

Studies the use of the iteratively reweighted least squares (IRLS) algorithm to obtain an L/sub p/-approximation for the design of two-dimensional (2D) linear-phase finite impulse response (FIR) digital filters. The algorithm is based on an iterative optimization technique first introduced in Burrus et al. for the design of 1D linear-phase digital filters and is very useful in the design of filters that give a trade-off between the traditional L/sub 2/ and L/sub /spl infin// designs. The new algorithm proposed, in terms of the maximum error in frequency response and complexity of implementation, compares favorably with the L/sub p/ optimization technique proposed by Lodge and Fahmy in [1980] and also with those obtained using other methods. Design examples and a comparison summary are presented. >