Hybrid methods of particle swarm optimization and spatial credibilistic clustering with a clustering factor for image segmentation

Abstract

Hybrid methods of fuzzy clustering and particle swarm optimization (PSO) are important techniques for image segmentation. The spatial credibilistic clustering (SCC) shows better performance than traditional fuzzy clustering, because of the “typicality” represented by credibility memberships degree is much more accurate than the “sharing” represented by probability membership degree to characterize the relationships between pixels and classes of images. Current integrated patterns of fuzzy clustering and PSO haven't made full use of both advantages. Therefore, main integrated forms were investigated and uniformly modeled by taking SCC as example, then a new kind of integrated pattern and algorithm was put forth, which integrates evaluation functions and update equations by introducing a clustering factor. Segmentation experiments validate that the method has better performance on running time and segmentation quality. The presented integrated pattern can be generalized to other hybrid methods of fuzzy clustering and PSO.