Abstract：A novel method for image segmentation with fuzzy C-means clustering based on image patch is presented．We introduce the image patch into clustering segmentation，and propose an IPFCM method．We utilize the image patch to replace the single pixel in fuzzy clustering that can increase the difference between different classes，and improve the membership functions to make it unimodal．The experimental results show that our method has strong anti-noise and high segmentation precision，the image of the membership function and the ideal membership functions are very close．At the same time our method need not too much control parameters，and has greater reliability and adaptability．On the other hand，each cluster center is the expansion of a vector and the update formula is given．This will provide the full rationale for image segmentation with FCM by using a variety of image features．
2011, Vol. 1 