Adaptive Thresholding for Pap-Smear.
- Author:
Kyung Hee SEO
1
;
Ho Sung KIM
;
Baek Sop KIM
Author Information
1. Department of Computer Science, Sungshin@Women's University, Korea. hkim@cs.sungshin.ac.kr
- Publication Type:Original Article
- Keywords:
Adaptive Thresholding;
Thresholding;
Pap-Smear
- MeSH:
Classification;
Cytoplasm;
Masks;
Neck;
Uterus
- From:Journal of Korean Society of Medical Informatics
2001;7(1):117-124
- CountryRepublic of Korea
- Language:Korean
-
Abstract:
We can assume a histogram of uterus neck cytoplasm image has three peaks which is consisted of nucleus, cytoplasm and background. We proposed a method for extraction of adaptive thresholding value that is suitable to each various intensity distribution. First, the adaptive thresholding is divided into thresholding of cytoplasm area and nucleus area. The thresholding of cytoplasm area, utilizing whole histogram, extracts thresholding value by using histogram standard deviation which of recognized as a background for each histogram distribution. The classification of nucleus is various in size and has difficulty in precise image extraction because of great difference in intensity in a cell image when using whole histogram distribution. So we suggests 'local thresholding' . In the first place, by using optimal thresholding, we can find nucleus seed area as a mask, and get adaptive thresholding value correct to each histogram distribution by obtaining histogram for each mask. Comparing to other methods that use the same thresholding value for one image, this can effectively extract nucleus and cyotoplasm. Because 'local thresholding' decides most suitable thresholding value for each distribution and characteristics.