A New Algorithm for Computing Threshold Value Used to Image Segmentation

Abstract:

With the development of computer technology, digital image technology is also to grow rapidly. At present, the image processing has become a popular research field. Segmentation is a very important critical step in image processing tasks, which directly affects the subsequent analysis results. Cluster analysis belongs to the category of unsupervised classification and it is not depended on any prior knowledge of the category, so it usually serves as a basic tool to be used for the initial stage of image segmentation. In this paper, we propose a weighted fuzzy threshold algorithm based on fuzzy relations. This algorithm determines the fuzzy membership function according to the similar relationship between pixels, and emphasizes influence of sharpening features of Gaussian function. The new method overcome the drawbacks of clustering algorithms such as, sensitive to the initial value, need to pre-set number of categories. The numerical results show the effectiveness of the new method.

• Periodical: Applied Mechanics and Materials (Volumes 580-583)
• Pages: 316-319
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.580-583.316
• Online since: July 2014
• Authors: Guo Yan Chen, De Qun Liang, Ying Zhang, Hai Yan Xie
• Keywords: Degree of Membership, Image Segmentation, Threshold

References

[1] Zhiwei Ben, Yunjie Zhao, in: Computer Applications and Software, Vol. 27 (2010), p.11. (In Chinese).

[2] Juan Bu, Xiangyang Wang, Yifeng Sun, in: Journal of Image and Graphics, Vol. 13 (2008), p.1837. (In Chinese).

[3] Jun Tian, Yongxuan Huang, Yalin Yu, in: Pattern Recognition and Artificial Intelligence, Vol. 21 (2008), p.221. (In Chinese).

[4] Chaira T, Ray A K, in: Pattern Recognition Letters, Vol. 24 (2003), pp.1837-1844.

[5] Chaira T, Ray A K, in: Appl, Soft Comput, Vol. 8 (2008), pp.919-927.

[6] Guo X.J., Zhang H.H., Chang Z., in: ICIC Express Letters, Vol. 3 (2010), p.1059.