Skip to Main Content
Standard Fuzzy C-Means (FCM) clustering has been widely used as an effective method for image segmentation. It gained a huge popularity because it efficient, straightforward, and easy to implement. However, FCM is sensitive to initialization and is easily trapped in local optima. In this paper, several entranced models for FCM clustering were proposed, namely W_SS_FCM, LAWSSSFCM and HFCM, to promote the performance of standard FCM. The proposed algorithms merge partial supervision with spatial locality to increase conventional FCM's robustness. The proposed models includes integrating adaptive filtering as a preprocessing step to FCM, using Laws level masks to obtain weighted-sum image for clustering, and Integrating both spatial information with partial expert knowledge in the FCM model to formulate a new Hybrid FCM (HFCM) model. A comparison study was conducted to validate the proposed methods' performance applying well established measures. Evaluation was done on three datasets: Brain MR phantoms, real Brain MR images and chest CT scans. Experimental results show considerable improvement over standard FCM and other variants of the algorithm. It also manifests high robustness against noise attacks.