Hybridization of Particle Swarm Optimization with Unsupervised Clustering Algorithms for Image Segmentation

Unsupervised fuzzy clustering algorithms are one of many approaches used in image segmentation. The Fuzzy C-means algorithm (FCM) and the Possibilistic C-means algorithm (PCA) have been widely used. There is also the generalized possibilistic algorithm (GPCA). GPCA was proposed recently and is a general form of the previous algorithms. These clustering algorithms can be trapped to the local optimal solutions. Hence, optimization techniques are often used in conjunction with algorithms to improve the performance. Some of optimization techniques have been inspired by nature such as swarm behavior. Particle Swarm Optimization (PSO) is one such technique. In this paper, PSO heuristics were combined with FCM, PCA, and GPCA algorithms to improve the overall clustering accuracy of these algorithms. To test the improvement with the PSO, these algorithms were tested on images. The overall effect of adding unique PSO methods was a higher percentage of satisfactory image segmentations.