中国邮电高校学报(英文) ›› 2021, Vol. 28 ›› Issue (6): 65-81.doi: 10.19682/j.cnki.1005-8885.2021.1014

• Artificial Intelligence • 上一篇    下一篇

Enhanced kernel-based fuzzy local information clustering integrating neighborhood membership

Song Yue, Wu Chengmao, Tian Xiaoping, Song Qiuyu   

  1. 1. School of Communication and Information Engineering, Xi'an University of Posts and Telecommunications, Xi'an 710121, China
    2. School of Electronic and Engineering, Xi'an University of Posts and Telecommunications, Xi'an 710121, China
  • 收稿日期:2020-12-28 修回日期:2021-09-05 出版日期:2021-12-30 发布日期:2021-12-30
  • 通讯作者: 宋玥 E-mail:528279022@qq.com

Enhanced kernel-based fuzzy local information clustering integrating neighborhood membership

  1. 1. School of Communication and Information Engineering, Xi'an University of Posts and Telecommunications, Xi'an 710121, China
    2. School of Electronic and Engineering, Xi'an University of Posts and Telecommunications, Xi'an 710121, China
  • Received:2020-12-28 Revised:2021-09-05 Online:2021-12-30 Published:2021-12-30

摘要:

To enhance the segmentation performance and robustness of kernel weighted fuzzy local information C-means (KWFLICM) clustering for image segmentation in the presence of high noise, an improved KWFLICM algorithm aggregating neighborhood membership information is proposed. This algorithm firstly constructs a linear weighted membership function by combining the membership degrees of current pixel and its neighborhood pixels. Then it is normalized to meet the constraint that the sum of membership degree of pixel belonging to different classes is 1. In the end, normalized membership is used to update the clustering centers of KWFLICM algorithm. Experimental results show that the proposed adaptive KWFLICM ( AKWFLICM) algorithm outperforms existing state of the art fuzzy clustering-related segmentation algorithms for image with high noise.

关键词:  image segmentation, fuzzy clustering, combined membership degree, local information factor

Abstract: To enhance the segmentation performance and robustness of kernel weighted fuzzy local information C-means (KWFLICM) clustering for image segmentation in the presence of high noise, an improved KWFLICM algorithm aggregating neighborhood membership information is proposed. This algorithm firstly constructs a linear weighted membership function by combining the membership degrees of current pixel and its neighborhood pixels. Then it is normalized to meet the constraint that the sum of membership degree of pixel belonging to different classes is 1. In the end, normalized membership is used to update the clustering centers of KWFLICM algorithm. Experimental results show that the proposed adaptive KWFLICM ( AKWFLICM) algorithm outperforms existing state of the art fuzzy clustering-related segmentation algorithms for image with high noise.

Key words:  image segmentation, fuzzy clustering, combined membership degree, local information factor

中图分类号: