Fuzzy local means clustering segmentation algorithm for intensity inhomogeneity image

Segmentation for images with intensity inhomogeneity is very difficult. In this paper, a fuzzy clustering-based method to segment intensity inhomogeneity images is presented. Firstly, a new expression of the fuzzy C-means(FCM) object function is derived through altering the prototype of every clustering to a point-wise function. Then, a weight function defined on the local window is introduced into the objective function. The local weight makes the prototype for every pixel depends only on the information of its local region, which is more reasonable for the considered problem. The proposed method has been applied to artificial and real-world images, e.g. X-ray vessel images and MRI brain images. The comparison segmentation results have shown the proposed model is very applicable for image segmentation with intensity inhomogeneity.