中文

Journal of Applied Sciences ›› 2009, Vol. 27 ›› Issue (3): 260-265.

• Signal and Information Processing • Previous Articles     Next Articles

Two-Stage Texture Image Segmentation

  

  1. 1. School of Communication and Information Engineering, Shanghai University, Shanghai 200072, China
    2. Institute of Intelligent Information Processing, Xidian University, Xi’an 710071, China
  • Received:2008-09-01 Revised:2008-10-08 Online:2009-05-28 Published:2009-05-28
  • About author:Corresponding Authors MA Xiu-li, Ph.D., lecturer, research interests including image processing, pattern recognition and intelligent information processing, E-mail: xlma@mail.shu.edu.cn; JIAO Li-cheng, Ph.D., professor, research interests including pattern recognition and intelligent information processing; WAN Wang-gen, Ph.D., professor, research interests including computer graphics, data visualization and data mining, E-mail: wanwg@staff.shu.edu.cn
  • Supported by:

    Project supported by the“863”National High-Tech Research and Development Program of China (No. 2007AA01Z319 );  the Innovation Foundation of Shanghai University (No. A.10-0107-07-005); the Research Foundation for the Excellent Youth Scholars of Higher Education of Shanghai (No. B.37-0107-07-702); the Shanghai’s Key Discipline Development Program (No. J50104)

PDF

1689

Cited

8

Abstract:

        Spectral clustering is a new clustering algorithm based on graph and similarity. For a very large image, it takes a long time to compute the affinity matrix, eigenvalues and eigenvectors. To apply the spectral clustering algorithm to large-scale clustering problems, a two-stage texture segmentation algorithm is proposed. An  improved watershed algorithm is used to perform pre-segmentation, followed by multiway spectral clustering with eigenvaluescaled eigenvectors to complete the segmentation. To verify the proposed algorithm, we apply it to texture image segmentation with satisfactory results.

Key words: watershed ,  multiway spectral clustering , feature extraction ,  texture image segmentation

CLC Number: