An Adaptive Image Segmentation Method Based on Kernel FCM Algorithm

Abstract

A new image segmentation method is proposed in this paper for improving the effect of the image segmentation. First, an original image is nonlinear mapped into a higher dimension kernel space, and the data are better separated under the kernel space comparing with that under the original image space, then, the number of categories of the image is determined by analyzing the image histogram using gauss filter method, and the detected peak point is as the initial center of the kernel fuzzy c-means (FCM) algorithm simultaneously, last, the kernel FCM algorithm is used to perform feature cluster. According to the results of experiments, the new image segmentation method which can adaptively achieve image segmentation and get better segmentation results, has stronger robustness to deal with the noise and the out-layer data comparing with the conventional FCM algorithm, and the analysis of image's histogram applied in the kernel FCM algorithm can greatly reduce the computational load.