Universal Nonlocal Fuzzy C-means for Image Segmentation

As an active tool for image segmentation, the classical fuzzy c-means (FCM) clustering algorithm is sensitive to noises. To overcome the weakness, several improved versions of FCM have been developed. Compared with the original one, these methods are more robust to noises. However, they are still not suitable for complex noises. To get better image segmentation results, especially in complex noisy scenarios, we provide a universal nonlocal FCM algorithm (UNFCM) by regularizing the objection function with a universal nonlocal spatial term. The experimental results on synthetic and real images suggest that the proposed UNFCM can yield higher quality segmentation results both in classification accuracy and visual effects.