基于支持向量机的肤色检测  

Skin detection based on support vector machines

作　　者：徐从东[1] 罗家融[1] 肖炳甲[1] 舒双宝[1] 

机构地区：[1]中国科学院等离子体物理研究所计算机应用研究室

出　　处：《计算机工程与设计》2008年第10期2600-2603,共4页Computer Engineering and Design

基　　金：中国科学院知识创新工程重要方向性基金项目(KJCX3.SYW.N4)

摘　　要：根据支持向量机理论和肤色信息分布特点,提出用支持向量机方法进行肤色检测。在YCbCr、YIQ、YUV颜色空间中,去除照度分量,分别以各像素点的色度向量为输入,用像素点所属类别——皮肤区域与非皮肤区域为输出,建立各颜色空间的支持向量机肤色检测器。实验表明,该方法具有良好的肤色检测效果。According to the theory of support vector machines and the characteristic of skin color distributing, a novel skin detection approach based on support vector machines is proposed. In YCbCr, YIQ and YUV color spaces, skin detectors based on support vector machines are constructed. Input vector of them is composed of pixers chrominance components except for luminance, and output is kinds including skin and nonskin which pixels belong to. Experimental result shows the approach to detect skin is effective.

关 键 词：肤色检测 支持向量机 颜色空间 高斯函数 序列最小最优化算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]