一种基于K-均值聚类的CIELAB和HSI色彩空间舌质舌苔分割算法  

Tongue body and tongue coating separation based on K-means Cluster in CIELAB and HSI color space

作　　者：郭宙[1] 杨学智[1] 司银楚[1] 朱庆文[1] 牛欣[1] 沙洪[2] 

机构地区：[1]北京中医药大学人体机能系,北京100029 [2]中国医学科学院生物医学工程研究所,天津300192

出　　处：《中华中医药杂志》2010年第5期663-665,共3页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家十一五科技支撑计划(No.2007BAIA107A23);国家自然科学基金重点项目(No.30672583)~~

摘　　要：目的:建立一种基于CIELAB(国际照明委员会1976年L*a*b*标准)和HSI(hue,saturation and intensity,色调、饱和度、亮度)色彩空间的舌质舌苔的分割算法。方法:以数码相机采集神经内科患者的舌图(有淡红舌薄白苔、腻苔、黄苔、舌红少苔、舌面瘀斑等舌图种类)为研究对象。算法结合了CIELAB和HSI色彩空间描述舌图的优势,先以舌图的a*值为聚类样本,使用K-均值聚类对舌质舌苔进行初步分割,然后再根据H值排除紫斑区域以完成分割。结果:该算法不仅可以分割舌质和舌苔,并且能识别舌质上的紫斑。使用该算法对56例采集自临床舌图进行舌质舌苔分割,成功率达92.85%。结论:该算法能自适应分割舌质舌苔,分割结果具有临床诊断价值。Objective：To establish a method for tongue body and tongue coating separation based on CIELAB and HSI color space.Methods：Pictures of tongue manifestation （including light red tongue and thin white tongue coating,greasy tongue coating,yellow tongue coating,red tongue and less tongue coating,ecchymosis on tongue body,etc） used in this study were get from patients of department of internal neurology by digital camera.This algorithm takes advantage of CIELAB and HIS color system to describe tongue image nature.First,K-means Cluster analysis method was used to separate tongue body and tongue coating by measure the a＊ value of every pixel.Then excluded the purple part of tongue body which was included in tongue coating.Results：The method was applied on 56 clinical tongue images with 92.85% of the test images successfully separated.Conclusion：The method can separate tongue body and tongue coating automatically and was meaningful for clinic diagnosis.

关 键 词：舌诊 CIELAB HSI 舌质舌苔分割 K-均值聚类 

分 类 号：R241[医药卫生—中医诊断学]