A Levinson-type algorithm for two-dimensional Wiener filtering using bivariate Szegö polynomials

Abstract

The Levinson algorithm has been widely utilized in the solution of the normal equations which arise in Wiener filtering. The algorithm can be easily derived and implemented using Szegö polynomials. In recent years, Wiener filtering techniques have been extended to two-dimensional processes. We show that the Szegö polynomials have bivariate analogues which can be used to derive and implement a two-dimensional analogue of the Levinson algorithm.