Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6901327 | Procedia Computer Science | 2017 | 8 Pages |
Abstract
Fuzzy c-means algorithm (FCM) is a powerful clustering algorithm and it is widely used in image segmentation. However, in FCM, we need to give both the parameters of the number of clusters and the initial membership matrix in advance, and they affect the clustering performance heavily. In this paper, we propose a novel density based fuzzy c-means algorithm (D-FCM) by introducing density for each sample. The density peaks are used to determine the number of clusters and the initial membership matrix automatically. Experimental results on benchmark datasets and medical image segmentation datasets show the efficiency and effectiveness of our D-FCM.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science (General)
Authors
Hua-Xin Pei, Zeng-Rong Zheng, Chen Wang, Chun-Na Li, Yuan-Hai Shao,