Design of image enhancement filters using a novel parallel particle swarm optimisation algorithm

Abstract

Designing image enhancement filters with arbitrary frequency response subject to stability constraints is a complex multidimensional optimisation problem. In this paper a novel parallel particle swarm optimisation algorithm (PPSO) is proposed and applied to the design of infinite impulse response image filters. The proposed PPSO consists of two phases: in the first phase, the particle swarm in the classical PSO algorithm is divided into subpopulations that evolve on separate cores of a multi-core machine. Best solutions from each sub population are then interchanged between cores. In the second phase a local search using Nelder-Mead simplex is done to refine the solution. Classical PSO is used for global exploration to explore multiple local minima whereas Nelder-Mead helps refine the solution computed by the PSO. The PPSO outperformed other global optimisation algorithms in terms of the mean square error between the ideal and designed filter frequency responses and CPU usage.