ABSTRACT

The fuzzy c-means clustering algorithm is one of the most important procedures of unsupervised clustering. It had always involved inadequacy for the necessity of determining the initial quantities and lack of ways on theoretical bases for choosing it, in view of the fact that in fuzzy c-means algorithm like unlinear optimizing procedures of initial quantities(like the cluster number and initial center and membership matrix) is efficient in convergence of this algorithm. So, if these quantities are not selected correctly for a noisy data, this algorithm will stop in local minimum. Therefore this algorithm becomes limit. This work develops a specific method to construct the initial cluster center to construct initial membership to clusters in order to improve the strength of the clusters. The algorithm is realized by incorporating the spatial neighborhood information to calculate the initial cluster center. So this paper finds a reasonable way to get the initial cluster center to initialized membership matrix. The new FCM are tested on a set of dataset and then the application to the segmentation of real MRI image is presented and compared with the results obtained using FCM.