Frequency sampling design of 2-D FIR filters based on DEPSO

Frequency sampling is a classical method for 2-D FIR filters design, which can be subdivided into the uniform and non-uniform approaches. However, previous work mainly focused on reducing the algorithm complexity or seeking the alternative implementation approaches rather than optimizing the samples. In this paper, a method to refine the conventional frequency sampling method is proposed, where samples equally distributed on the Cartesian grid are adjusted by utilizing DEPSO algorithm so as to achieve the nearly optimum minimax design. Simulation results demonstrated that this new method outperforms the conventional methods on minimax sense. And the generated filters can give better image denoising performance than the median filter.