Design of 2-D multiplierless FIR filters using genetic algorithms

This paper considers the design of reduced complexity two-dimensional FIR filters using genetic algorithms (GAs). Circularly symmetric and diamond shaped low-pass linear phase FIR filters are designed using coefficients comprising the sum or difference of two power-of-two terms. A minimax error criterion is adopted which leads to a minimisation of the weighted ripple extrema in both pass and stop bands. The results presented are compared with those obtained using simulated annealing, linear programming and simple rounding of an optimum (continuous) minimax solution. The filters designed using GAs are found to be as good or better than those designed using other methods, with less computational effort than the simulated annealing approach