Subband domain Wiener filtering

A general FIR solution of subband domain Wiener filtering is derived. We observe some interesting properties of the optimal filters and the solution is generalized to the multi-scale subband domain which paves a way for the study of wavelet domain Wiener filtering. Unlike the unitary transform domain Wiener filtering, the performance of the general subband Wiener filter depends on the choice of the subband filter. An algorithm is derived for the identification of the optimal paraunitary analysis and synthesis filters which together with the subband Wiener filter, provides the minimum mean-square estimation error.