Color Image Region Growth Segmentation Integration of Normalized Cut

Abstract:

This paper solves that image segmentation result is not consistent with human visual perception or too broken. First of all, based on the continuity of image features, appropriate human vision, calculated the similarity of color image pixel as Eq.2 in HSV space to grow region, then made the regional merge, using normalized-cut segmentation method as Eq.4 and Eq.5 to eliminate over-segmentation phenomenon. In this paper, experimental results shows that the segmentation can be achieved very good results as Fig.1, and parts of the method can be applied in other segmentation to solve over segmentation. This method on color images as the research object is different from other methods on gray images, the selection of seeds and achieves these automatic that differ from general algorithms, presents a new implementation to solve over-segmentation.

• Periodical: Advanced Materials Research (Volumes 143-144)
• Pages: 139-142
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.143-144.139
• Online since: October 2010
• Authors: Xiao Ying Wu, Li Juan Ma, Zhao Feng Li, Shi Tao Yan
• Keywords: Color Image Segmentation, Normalized Cut, Region Growth

References

[1] ShiJ, Malik J: Normalized Cuts and Image Segmentation [J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, (2000).

DOI: 10.1109/34.868688

[2] Zhou Ming-Quan: Content-based image retrieval, Beijing. Tsinghai University Press . (2007).

[3] Zhang Yu-Jin: Image Engineering , image analysis Vol. Beijing. Tsinghai University Press . (2005).

[4] Mu Tao: A kind of integration of regional growth and graph theory image segmentation. Computer Engineering and Applications . (2005).

[5] Wang Ying and so on: A kind of integration of region growing and edge detection of color image segmentation. Technology Review . (2008).

[6] ShiJ, Malik J: Normalized Cuts and Image Segmentation[C]. Proc. Of IEEE Conf. Computer Vision and Pattern Recognition, (1997).