Journal of South China University of Technology (Natural Science Edition) ›› 2015, Vol. 43 ›› Issue (5): 59-65.doi: 10.3969/j.issn.1000-565X.2015.05.010

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

Edge Detection of Images Based on NSCT and KFCM

Wu Yi-quan1,2,3,4 Zhu Li1 Li Li1   

  1. 1． College of Electronic and Information Engineering，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，Jiangsu，China; 2． Jiangsu Provincial Key Laboratory of Pulp and Paper Science and Technology，Nanjing 210037，Jiangsu，China; 3． Key Laboratory of High-Speed Railway Engineering of the Ministry of Education，Chengdu 610031，Sichuan，China;4． Shenzhen Key Laboratory of Urban Rail Traffic，Shenzhen 518060，Guangdong，China
  • Received:2014-09-30 Revised:2015-01-27 Online:2015-05-25 Published:2015-05-07
  • Contact: 吴一全(1963-)，男，博士，教授，博士生导师，主要从事图像处理与分析、目标检测与识别、智能信息处理研究． E-mail:nuaaimage@163.com
  • About author:吴一全(1963-)，男，博士，教授，博士生导师，主要从事图像处理与分析、目标检测与识别、智能信息处理研究．
  • Supported by:
    Supported by the Foundation of Jiangsu Provincial Key Laboratory of Pulp and Paper Science and Technology(201313)，the Foundation of Key Laboratory of High-Speed Ｒailway Engineering of the Ministry of Education(2014-HＲE-01) and the National Natural Science Foundation of China(61203246，61102131)

PDF

218

Abstract: In order to improve the performance of existing image edge detection methods，a novel edge detection method on the basis of nonsubsampled contourlet transform (NSCT) and kernel fuzzy c-means clustering (KFCM) is proposed． In this method，firstly，an original image is decomposed into a low-frequency component and some high-frequency components via NSCT． Secondly，edge information is extracted from the low-frequency component with less noise and is clustered via KFCM to obtain low-frequency edge image． As a result，the accuracy of edge localization is improved． Then，in order to decrease pseudo-edges and richen image details，the method of modulus maxima is applied to high-frequency components with more edges and details． Finally，the whole image edge is obtained by fusing the edge images of low-frequency component and high-frequency components． Experimental results show that，in comparison with the Canny method，the method on the basis of edge detection operator and fuzzy clustering，the method on the basis of edge information and fuzzy c-means algorithm optimized by chaotic par-ticle swarm，as well as the method of modulus maxima in NSCT domain，the proposed method helps obtain better edge detection effect with accurate edge localization，continuous and complete edges，as well as abundant details．

Key words: image processing, edge detection, nonsubsampled contourlet transform, kernel fuzzy c-means cluste-ring, modulus maxima

CLC Number: