Comparison of Fuzzy Algorithms on Images

Image segmentation is a process by which an image is partitioned into regions with similar features. Many approaches have been proposed for image segmentation, but generally we use Fuzzy C-Means method, because it gives better results for large class of images. However, using this method is not suitable for images with noise and it is a lengthy process in terms of duration when compared with other method. For this reason, many other methods have been proposed to improve the shortcomings of image segmentation using fuzzy C-Means. Techniques like Credibilistic Fuzzy C-Means overcomes the problem of noise persisted using FCM. Intuitionistic Fuzzy C-Means introduces the concept of non-membership for a cluster. Krishnapuram and Keller [1] suggested usage of Possibilistic C-Means clustering which relaxes the column constraint of FCM so that membership matrix better reflects the typicality of particular data point in a cluster and noise could be avoided. We perform a comparison of these clustering algorithms on the basis of execution time and validity function for each algorithm applied on different kind of images taken in consideration.