摘要
研究了人脸检测中的肤色检测问题,针对广泛使用的YCbCr颜色空间存在受色差以背景干扰影响检测率的缺点,为提高人脸肤色的检测率和检测速度,提出了结合人脸肤色在RGB三个维度上的分配比例而对传统的YCbCr空间分量进行改进的方法,得出YCrCg平面,提取肤色分量,进而引入高斯密度函数估计和直方图统计的方法构建肤色模型,使肤色区域从非肤色区域当中准确地分离出来。在确定人脸区域范围之后,从该图像区域中提取出唇色信息,考虑到人脸的旋转角度,得到检测之后的人脸图像区域。通过自建式网络人脸肤色图库进行仿真,结果证明改进的肤色检测模式取得到较好的效果。
This paper analysis the skin color detection mode in Face Detection.For the current widespread use of YCbCr color space has the shortcoming of background color interference,in order to improve the skin detection face detection rate and detection speed,the paper proposes an improving approach which combined the RGB distribution ratio of the face color with the traditional components of YCbCr space,based on this,YCrCg plane was drawn,skin color component was extracted,and the Gaussian density function estimation and statistical methods were introduced to build the histogram color models so that the non - skin color region was accurately separated.After determining the scope of the face region,the lip color image information was extracted,and taking into account the rotation angle of the human face,the detected human face image area was obtained.Based on the human face color gallery of the simulation results show that the skin detection model has good results.
出处
《计算机仿真》
CSCD
北大核心
2010年第7期222-225,共4页
Computer Simulation
关键词
人脸检测
颜色空间
肤色分量
Human face detection
Color space
Color component
作者简介
黄秀常（1971．2-），男（汉族），浙江浦江人，工程硕士，讲师，主要研究领域为图形图像、多媒体。
