关键词: 关键词: 图像分割, 模糊C均值算法, 聚类算法, 空间信息, 隶属度

Abstract: Abstract: Fuzzy C-Means Clustering (FCM) as one of the clustering methods is often used for image segmentation, but the traditional FCM is sensitive to noise. Recently, many researchers have incorporated local spatial information into the original FCM algorithm to improve the performance of image segmentation. But, there is a crucial parameter in their objective functions, used to balance between robustness to noise and effectiveness of preserving the details of the image. Moreover, the value of the parameter has to be made by experience or trial and error experiments. Although the introduction of local spatial information enhances their insensitiveness to noise to some extent, they still lack enough robustness to noise and outliers. To overcome the problem，we propose a FCM extended algorithm which uses neighboring membership instead of the parameter. The proposed algorithm is completely free of the parameter that controls the balance between the image noise and the image details, which is automatically achieved by the definition of the fuzziness of each image pixel (both spatial and gray level). Experiments implemented on real-world images demonstrate that the proposed method achieves better performance for image segmentation, especially for images corrupted by intense noise, compared to the traditional FCM and its extended methods.

Key words: Keywords: Image segmentation, Fuzzy c-means clustering, Clustering algorithm, Spatial information, Membership function