Research Of Image Segmentation Based On Fuzzy Clustering Theory

Image segmentation is a kind of image processing methods, which is to split a image into some classes with the same propertied and extract the target area from these classes by relevant rules. Image segmentation algorithms based on fuzzy clustering theory have attracted widespread attention and application because of its good segmentation performance.Fuzzy c-means(FCM) clustering algorithm is a kind of unsupervised clustering algorithms and has good convergence. However, this method simply considers the gray scale information of image and ignores spatial information of the image, which will lead to the loss of image detail information, inaccurate segmentation result, noise interference and other issues. In this paper,the image segmentation method based on fuzzy theory is studied and the intensive researches of traditional FCM algorithm and its improved algorithm are made. To address these problems aforementioned, the work about FCM algorithm and its improved version in image segmentation are done as follow:(1)This paper focused on the intensive study of standard FCM image segmentation algorithm and its improved version, and the strengths and deficiencies of different algorithms are analyzed. Use image filtering based on FGFCM method that utilize this correlation to replace the constant in EnFCM algorithm, which can effectively balances the noise and image detail. Afterwards, KFCM algorithm that is based on kernel function is studied.The experimental results show that FGFCM is effective in balancing the noise and image details,the segmentation effect is relatively ideal.(2)Further research on RFCM algorithm is made. In order to effectively avoid noise interference caused by ignorance of the relationship between noise and neighborhood, a new fuzzy C means clustering image segmentation(SNRFCM) algorithm that is the combination of spatial neighborhood and grayscale information is proposed.(3)On the basis of SNRFCM algorithm, further study and analysis are made. Weproposed a kernel fuzzy C-means clustering image segmentation(SKNRFCM) technique based on spatial information by replacing Euclidean distance with kernel sensing distance in the objective function and introducing kernel function. Using a penalty factor can effectively reduce the negative effects of noise on the clustering result. And using cluster weights can effectively prevent noise pollution. Experimental analysis show that this proposed method has good segmentation performance especially in segmentation details and edges of image while having better anti-noise capacity when compared to FCM, RFCM, NRFCM, KGFCM and FGFCM.