基于统计的面部校色法  

Method of Face Modifing in Color Based on Statistics

作　　者：张华[1] 邵桂芳[2] 

机构地区：[1]西南科技大学信控学院,四川绵阳621002 [2]重庆大学自动化学院,重庆400044

出　　处：《绵阳经济技术高等专科学校学报》2002年第1期46-50,共5页Journal of Mianyang College of Economy & Technology

摘　　要：在虚拟照像系统中 ,肤色特征易受光线、背景色彩和观测角度的影响 ,如有强光时人的脸色就偏暗 ,严重影响了视觉效果。提出了采用软件技术来加以补偿的方法。该方法是基于概率统计和模糊统计知识 ,根据基于肤色的人脸实时跟踪技术[2 ] 来建立肤色模型 ,通过Matlab工具软件来进行实验训练 ,提取出最佳肤色特征值作为参考来对整个图像进行识别 ,从而检测到人的面部 ,进而对其进行校色处理 ,有效地保证了虚拟照像系统的相片质量 ,最后我们利用VisualC + +实现该算法。实践表明 。In the VR photo system, since face color is so liable to be affected by light?background color and observed angle, such as peoples' face color will be dim when the light is strong, that it badly influence the sight effect. Owing to this we put forward a supplement method based on software technology, during the exploition of the theme. The method has effectly ensured the quality of the photograph of VR photo system ,which bases on the knowledge of probability statistics and fuzzy statistics, builds the skin-color model according to the technology of skin-color based face tracking in real-time, goes on a series of training by use of Matlab toolbox in order to find the just characteristics to recognize the whole image , finally finds the face and deal it with color. At last we make use of Visual C++ to realize the method.

关 键 词：虚拟照像系统 概率统计 模糊统计 肤色模型 面部校色法 虚拟现实 图像识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP391.9[自动化与计算机技术—计算机科学与技术]