A coevolutionary approach for optimization of image enhancement filters

In this paper we demonstrate how coevolutionary algorithms (CEAs) can be employed for optimization of image enhancement filters. Specifically, we take the example of an impulse noise filter constructed using the fuzzy paradigm, and show how a CEA could be effectively used for its optimization. The fuzzy impulse filter we take results from our research where it was seen that the shape and the corresponding parameters of the membership functions used in the fuzzy inference process, play a major role in the quality of the enhanced image, apart from the proper selection of the fuzzy rule base. This is true, both in terms of objective and subjective evaluations of the processed image. During our experiments, we employed a CEA (having a blend of cooperativeness and competitiveness) to optimize the rule base and to select the best shape of the membership function used in the fuzzy inference process, and we present the results for several real images, to show the effectiveness of the proposed approach.