基于多关联模板匹配的人脸检测

提出一种基于多关联模板匹配的人脸检测算法.模板由一系列关联的双眼模板和人脸模板组成,它们都是通过仿射变换根据伸缩比和姿态(即旋转角度)从单一平均脸模板产生出来的.首先,使用双眼模板搜索候选人脸,再用人脸模板匹配进一步筛选候选人脸,最后,通过启发式规则验证是否是人脸.对于各种类型的图像进行大量实验的结果表明，该算法对于正面包括多角度人脸的检测很有效.

Face Detection Based on the Matching of Multiple Related Templates

In this paper, a face detection algorithm based on the matching of multiple rela ted templates is presented. The templates are a series of related types: eyes-i n-whole and face itself, which are produced by affine transforms varying in str etch and pose from an average frontal face. The eyes-in-whole templates are us ed as the first search step for face candidates and then the face templates are matched. Finally some heuristic rules are used to verify face detection results. Experimental results obtained from images containing frontal and in-plane rota ted faces demonstrate the feasibility of this approach.