Hybrid Fuzzy C-Means Using Particle Swarm Optimization (PSO) and Differential Evolution (DE) for Image Segmentation

Among fuzzy clustering methods, fuzzy C-means (FCM) have become more popular in image segmentation because of their simplicity and effectiveness. However, FCM is generally very sensitive to noise and highly dependent on the initialization of cluster centers, making it fall into local maxima. In this paper, a hybrid FCM algorithm (HDEPSO-FCM) that integrates particle swarm optimization (PSO), and differential evolution (DE) is proposed to address the weaknesses of FCM. The idea of hybridizing PSO and DE into FCM has refined the segmentation quality by providing anti-noise ability. In addition, HDEPSO-FCM initializes the cluster centers optimally, thus, faster segmentation. Hence, HDEPSO-FCM performs relatively better than other state-of-the-art segmentation methods when tested with the Berkeley Segmentation Data Set (BSDS) images. Remarkably, out of 500 images, HDEPSO-FCM has obtained a better average PRI of 0.7480 when compared to other techniques. Moreover, the computational cost of HDEPSO-FCM to segment 100 test images is also faster by 43.8%. Thus, the proposed HDEPSO-FCM algorithm is an effective alternative method in image segmentation.