摘要
通过对图像进行预处理 (似眼物检测 ) ,不但缩小了人脸检测的搜索空间 ,而且将复杂背景下多姿态多人脸检测简化为在候选区域内 ,已知大致的人脸尺度情况下正面端正单人脸检测问题 ,并结合多模板匹配与规则验证完成了人脸区域的最后确定
By using the eye-analog detection, the algorithm can not only reduce the search space of face detection, but also simplify the multi-pose multi-face detection in a complicated background into single front-view vertical face detection in the candidate regions. With knowing the approximate size of the face, it uses multi-template matching and rule verification to determine the location of real face. Experiments demonstrate that it is relatively fast and rather successful.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2004年第1期45-50,共6页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金 (60 13 5 0 10 )资助
关键词
人脸检测
多模板匹配
图像分割
图像处理
似眼物检测
face detection
eye-analog detection
maximum entropy image segmentation
template matching
