融合区域颜色和纹理两级特征的快速人体皮肤检测  

RAPID HUMAN SKIN DETECTION WELDING REGIONAL COLOUR AND TEXTURE TWO-LEVEL CHARACTERISTICS

作　　者：刘忠平[1] 邵平[1] 

机构地区：[1]玉林师范学院职业技术学院,广西玉林537000

出　　处：《计算机应用与软件》2010年第10期134-137,共4页Computer Applications and Software

基　　金：广西教育厅项目(200911LX357)

摘　　要：人体皮肤检测在人脸检测、成人图像过滤、人体图像检索等应用中占有重要地位。利用支持向量机,对人体皮肤的颜色和纹理特征的分布进行了研究,并提出了一个基于区域颜色和纹理特征规则的两级模型。在人体皮肤检测算法中,首先利用分水岭分割算法将图像分割成颜色和纹理近似一致的区域,然后利用皮肤颜色模型提取候选皮肤区域,最后利用纹理规则模型对候选皮肤区域进行最终判决。实验结果表明,该算法简便、快速、有效。Human skin detection plays an important role in face detection,adult image filtering and human image retrieval applications and so forth.By using SVM this article investigates into the characteristic distributions of human skin colour and texture and presents a two-level model based on regional colour and texture rules.According to the skin detection algorithm an image is firstly segmented into homogeneity regions according to colour and texture by the watershed method.Then the skin region candidate is extracted by the skin colour model.Finally the skin region candidate is judged by the texture rule model.Experimental results show that the algorithm is convenient,rapid and efficient.

关 键 词：皮肤检测 颜色模型 纹理模型 图像分割 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]