摘要
构造了一个彩色图片的正面人脸检测系统。首先利用肤色在YCbCr空间中沿Y方向的集中分布特性构建肤色信息库,根据该信息库在图像中检测出肤色区域;然后在肤色区域利用贝叶斯特征判别方法进行正面多尺度人脸检测。另外,定义了一些启发式搜索规则,有效地加快了人脸目标的搜索速度。实验证明,用较少的样本进行训练的人脸检测系统,对有复杂背景、多样化的测试集具有较好的测试效果。
A frontal face detection system in color images was constructed. First, a skin color information database was established according to the characteristic that skin color is distributed in centralized region along Y axis in YCbCr space. According to the database, skin region could be detected. Second, the Bayesian statistical model using discriminating feature was constructed to detect multiple frontal faces in skin region. Finally, in order to accelerate the face searching, some heuristic searching criteria were defined. The experimental results show that the face detection system uses few of samples, but displays relatively better performance while testing images with complicated background and diverse sources.
出处
《计算机应用》
CSCD
北大核心
2006年第8期1854-1856,1859,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60021302)
国家863计划资助项目(2005AA1470602005AA1Z1272)
关键词
人脸检测
肤色模型
贝叶斯待征判别方法
启发式搜索策略
face detection
skin color model
Bayesian Discriminating Feature(BDF) method
heuristic searching criteria