Multi-objective fuzzy clustering method for image segmentation based on variable-length intelligent optimization algorithm

Abstract

Fuzzy clustering method for image segmentation usually needs the determination of the cluster number in advance. Therefore, an adaptive fuzzy clustering image segmentation algorithm based on jumping gene genetic algorithm (JGGA) is investigated in this paper. A new weighted multi-objective evaluation function considering the cluster number, the inner-class distance and the inter-class distance is proposed. Because the cluster number is uncertain during the optimization process, a variable-length JGGA (VJGGA) is introduced. The cluster number and the cluster centers of the image gray values are determined by the minimization of evaluation function. Simulation results of the segmentation for a real image indicate that VJGGA algorithm is characterized by strong global capability of searching the optimal segmentation number and cluster centers, compared with variable-length GA (VGA). © 2010 Springer-Verlag Berlin Heidelberg.