基于贝叶斯判决的关于YCbCr空间的肤色模型查询表建立的研究

皮肤颜色信息是人脸检测和定位中一种有效的健壮特征.肤色模型的优劣主要取决于在肤色与非肤色交叠区域建立判决机制的有效性.提出了一种基于贝叶斯判决提取YCbCr颜色空间数据的肤色分类方法.该方法从大量肤色样本和非肤色样本统计出实验数据,使用贝叶斯规则建立肤色分类器.考虑到在高亮度区和低亮度区肤色色度(Cb,Cr分量)与亮度(Y分量)非线性相关,采用Y-Cb和Y-Cr两个子空间的查询表来建立肤色模型.实验表明,引入贝叶斯判决可以通过改变阈值来调节漏检率和误检率变化;采用两个查询表比传统的Cb-Cr查询表有更好的检测效果.

A Study of Building Skin Color Model of Lookup Table in YCbCr Color Space Based on Bayes Decision

Skin color information has proven to be a useful and robust character for face detection and localization. Whether a skin color model is good or bad mainly depends on its classification rule of discriminating skin and non-skin color overlap. A skin color classification technique that employs Bayesian decision with color statistics data gathered from YCbCr color space was presented. By using Bayesian rule the experimental data can be processed from a great amount of skin color and non-skin color samples to establish skin color classifier. Considering the non-linearity among chroma (Cb, Cr component) and luminance (Y component) of skin color in the high and low luminance region, the skin color model is constructed by two lookup tables in Y-Cb and Y-Cr subspaces. Experimental results have demonstrated that the amount of false detection and false dismissal can be controlled by adjusting the threshold value and two lookup tables (LUT) can achieve better classification outcomes than traditional Cb-Cr lookup table.