Image guided fuzzy clustering for image segmentation

Abstract

Fuzzy clustering methods are efficient tools for image segmentation. However, most of fuzzy clustering approaches are too sensitive to deal with the misclassification of pixels in image segmentation. In recent years, a variety of enhanced fuzzy clustering approaches have been proposed to obtain smoother results in noised image segmentation, but usually with less accurate edges in these results. To fix this problem, we derive some modified algorithms by using Guided Filter, the filter that can reserve the edge information when smoothing every region in segmentation. This paper provides a new roadmap for the application of Guided Filter and gives a thorough discussion of its applications in classical clustering methods, which are Fuzzy C-Means (FCM) and some other variants of FCM. Verified by the experimental results, we conduct a good use of Guided Filter to improve the performance of fuzzy clustering methods in a simple way.