摘要
提出了一种基于肤色分割和模板匹配相结合的人脸检测算法。首先利用rgb色彩空间下的人脸肤色模型,对人脸图像进行肤色分割;针对图像中存在的多个肤色区域连接在一起的问题,采用SUSAN算子提取区域的边界,将连接的肤色区域分开;根据肤色区域的形状特征和欧拉数筛选人脸候选区域;最后利用建立的人脸模板和一种改进的混合匹配准则,对候选人脸区域进行匹配识别。实验结果表明,该方法能较好地从复杂背景中检测出人脸。
A face detection algorithm is proposed which is based on combination of complexion segmentation and template matching.First,the face complexion model in rgb colour space is used to segment complexion of face image.For the issue existing in the image that many complexion regions connect together,the SUSAN operator is used to extract the boundaries of the region to separate the connecting complexion regions.Human face candidate regions are screened according to the shape feature of the complexion region and the Euler number.At last,the established human face templates and an improved mix-match criterion are used to identify candidate face regions.Experimental results show that the method can well detect human faces from complex background.
出处
《计算机应用与软件》
CSCD
2011年第7期112-114,140,共4页
Computer Applications and Software
基金
江苏省高校自然科学基金(BK2009116)
