关键词: FCM, 内核诱导距离, 核聚类, 鲁棒性

Abstract:

When Traditional FCM algorithm processes the noisy images,it exists the poor performance of image denoising,slow clustering,and poor segmentation performance.The nuclear clustering algorithm is combined with the traditional FCM algorithm using the kernel induced distance instead of the new clustering method of Euclidean distance.a modified kernel-based FCM clustering algorithm is deduced by using the spatial information and features of samples.Comparative tests show that this algorithm reduces the time of image segmentation and de-noising by 68% with an improved PSNR by about 10% compared with the traditional FCM algorithm,exhibiting better noise resistance and robustness.

Key words: FCM;kernel induced distance;kernel clustering;robustness