电子学报 ›› 2017, Vol. 45 ›› Issue (7): 1695-1700.DOI: 10.3969/j.issn.0372-2112.2017.07.020

• 学术论文 • 上一篇    下一篇

基于学生t分布的鲁棒分层模糊算法及其在图像分割中的应用

徐超1, 詹天明1, 孔令成2, 张辉1,2   

  1. 1. 南京审计大学工学院, 江苏南京 211815;
    2. 南京信息工程大学计算机与软件学院, 江苏南京 210044
  • 收稿日期:2016-11-09 修回日期:2017-02-27 出版日期:2017-07-25 发布日期:2017-07-25
  • 作者简介:徐超,男,1980年出生,湖北红安人,博士,副教授,研究方向为可信软件、软件工程和嵌入式系统.E-mail:xuchao@nau.edu.cn;詹天明,男,1984年出生,江苏省高邮人,博士,副教授,研究方向为模式识别、图像处理与大数据分析.E-mail:ztm@nau.edu.cn
  • 基金资助:

    国家自然科学基金(No.61572257)(No.61640220);江苏省高校自然科学研究重大项目(No.16KJA52002);江苏省"六大人才高峰"高层次人才资助项目(No.2015-XXRJ-015);南京审计大学政府审计重点项目(No.D2010530068)

A Robust Hierarchical Fuzzy Algorithm with Student's t-distribution for Image Segmentation Application

XU Chao1, ZHAN Tian-ming1, KONG Ling-cheng2, ZHANG Hui1,2   

  1. 1. School of Technology, Nanjing Audit University, Nanjing, Jiangsu 211815 China;
    2. School of Computer & Software, Nanjing University of Information Science & Technology, Nanjing, Jiangsu 210044 China
  • Received:2016-11-09 Revised:2017-02-27 Online:2017-07-25 Published:2017-07-25

摘要:

著名的模糊C均值算法(FCM)一直被视为图像分割应用中一个强有力的工具.然而,由于FCM中距离函数选择问题使得其对图像噪声的鲁棒性不足.本文提出了一个新的分层模糊C均值算法,使得传统的模糊C均值算法对于图像噪声和离群点有更好的鲁棒性.在此基础上引入了一个更加灵活的函数,即将距离函数本身看作是一个子学生t分布函数.使分层模型具有更好的通用性和灵活性.本文提出的算法可以扩展到其他基于FCM模型的算法实现,以获得更优的鲁棒性.实验结果表明本文提出新的分层模糊C均值算法的鲁棒性确实有效.

关键词: 分层算法, 模糊C均值, 图像分割, 学生t分布

Abstract:

The well-known fuzzy c-means algorithm (FCM) has been regarded as a useful tool for image segmentation application.However,it is still insufficient robustness to image noise due to the distance function selection in FCM.In this paper,we propose a new hierarchical fuzzy algorithm to make the traditional fuzzy c-means more robust to image noise and outliers.We introduce a more flexibility function which considers the distance function itself as a sub-FCM with student's t-distribution.Thus,our hierarchical model is general and flexible enough to deal with outliers and noises.Our algorithm proposed in this paper can be extended to any other FCM-based models to achieve superior performance.Experimental results demonstrate the improved robustness and effectiveness of the proposed algorithm.

Key words: hierarchical algorithm, fuzzy c means, image segmentation, student's t-distribution

中图分类号: