Two-stage morphological filter design using genetic algorithm

Morphological filters are an important class of nonlinear signal/image processing and analysis tools. These filters have been successfully used in a wide range of applications. Designing this kind of filters needs a prior knowledge in mathematical morphology. Genetic algorithm is an automatic approach to design of these filters which needs little prior knowledge. In this paper, we propose some heuristics to improve the designed filters such as integer representation instead of binary representation in chromosomes, using four operators instead of the conventional two, and using two structural elements instead of one. Experimental results with some examples in noise reduction tasks are shown.