A Novel Robust FCM Algorithm Combining Local Information on Edge for Image Segmentation

The standard fuzzy c-means (FCM) for image segmentation is very sensitive to noise and cannot get weak boundary accurately due to its insufficient usage of local information.To overcome these drawbacks, we present a novel image segmentation algorithm called GLFCM by combining local information on edge with FCM algorithm.First, the algorithm designs a new fuzzy factor integrating coefficient of variation of local windows and pixel gray value similarity, to control the trade-off between smoothing and clustering more accurately and enhance the robustness.Then, local windows near boundaries are selected and local information is utilized to optimize the pixels near the boundaries.Finally, we segment the selected windows, to improve the incorrect parts near the boundaries of the clusters.Experimental results show that the proposed algorithm has the stronger anti noisy property and higher segmentation accuracy.