2-D FIR eigenfilters: a least-squares approach

The 1-D eigenfilter approach is extended to the design of 2-D FIR (finite-impulse-response) filters. By minimizing a quadratic measure of the error in the 2-D frequency band, an eigenvector of an appropriate matrix is computed to obtain the filter coefficients. The method is not only simple but also is optimal in the least-squares sense. Several numerical design examples for arbitrarily shaped 2-D filters are used to illustrate the effectiveness of the approach.>