11 January 2007 New 2D adaptive image thresholding method based on within and between cluster scatter
Author Affiliations +
Proceedings Volume 6279, 27th International Congress on High-Speed Photography and Photonics; 62793N (2007) https://doi.org/10.1117/12.725364
Event: 27th International congress on High-Speed Photography and Photonics, 2006, Xi'an, China
Abstract
In this paper, a novel image thresholding approach based on within and between cluster scatter of the 2D histogram is proposed. 2D Otsu's method not only utilizes the gray-level information of each pixel but also synthesizes its spatial correlation information within the neighborhood, so it is more valid in the threshoding. However, it doesn't take into account of the within cluster scatter, sometimes it can't reach the optimal thresholding validity. In this approach, within and between cluster scatter are integrated to improve the algorithm. The threshold vector that maximizes between cluster scatter and minimizes within cluster scatter at the same time is regarded as optimal threshold. In order to avoid exhaustively computing within cluster scatter, we simplify the threshold recognizing function via replacing within cluster scatter by the difference in total covariance and between cluster scatter. As the total covariance of a given image is a certain constant, the simple algorithm will decrease complexity of computation greatly without affecting the segmented result. Experimental results show that our proposed approach not only performs well and effectively but also is more robust when applied to noisy image.
© (2007) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Yunyan Zhou, Yunyan Zhou, Kuntao Yang, Kuntao Yang, } "New 2D adaptive image thresholding method based on within and between cluster scatter", Proc. SPIE 6279, 27th International Congress on High-Speed Photography and Photonics, 62793N (11 January 2007); doi: 10.1117/12.725364; https://doi.org/10.1117/12.725364
PROCEEDINGS
7 PAGES


SHARE
Back to Top