Image Clustering Using Improved Particle Swarm Optimization

International audienceIn this paper, we propose an improvement method for image segmentation problem using particle swarm optimization (PSO) with a new objective function based on kernelization of improved fuzzy entropy clustering algorithm with spatial local information, called PSO-KFECS. The main objective of our proposed algorithm is to segment accurately images by utilizing the state-of-the-art development of PSO in optimization with a novel fitness function. The proposed PSO-KFECS was evaluated on several benchmark test images including synthetic images (http://pages.upf.pf/Sebastien.Chabrier/ressources.php), and simulated brain MRI images from the McConnell Brain Imaging Center (BrainWeb (http://brainweb.bic.mni.mcgill.ca/brainweb/)). Experimental results show that our proposed PSO-KFECS algorithm can perform better than the competing algorithms