Segmentation of tongue body and fur based on common color spaces of K-means clustering
10.3321/j.issn:1006-2157.2009.12.007
- VernacularTitle:基于K-均值聚类的常用色彩空间舌质舌苔分割研究
- Author:
Zhou GUO
1
;
Xue-Zhi YANG
;
Yin-Chu SI
;
Qing-Wen ZHU
;
Xin NIU
;
Hong SHA
Author Information
1. 北京中医药大学
- Keywords:
tongue diagnosis;
color space;
segmentation of tongue body and fur;
K-means clustering
- From:
Journal of Beijing University of Traditional Chinese Medicine
2009;32(12):819-821,后插1
- CountryChina
- Language:Chinese
-
Abstract:
OMective To observe which pathways(or components)of common color spaces are good for discriminating and segmenting tongue body and fur.Methods The typical tongue image(pink tongue with thin-white fur.greasy fur and yellow fur)and lower color temperature tongue image were taken as study objects,and special color component of every pixel of tongue imagines(value of a~*,L~*,H and R)in the color spaces of red,green and blue(RGB),L~* a~* b~* criterion of International Commission on Illumination in 1976(CIELAB),and hue,saturation and intensity(HSI)were selected as clustering objects.K-means clustering was applied to generate two clusters and complete the segmentation according to the characteristics of them.Results The Component of a~* value of CIELAB color space was the best in the aegment effect and anti-interference capability,which was not influenced by illumination angle,bias of color temperature and types of tongue body and fur.The clustering resuhs of L*value and R value were influenced by uneven illumination significantly.The clustering result of H value showed a state of incomplete segmentation.Condusion The segmentation of tongue body and fur had a stable result taking a~* value in CIELAB color space based K-means clustering as index and showed practical values in tongue diagnosis.
