基于家庭数字相册的人脸检测与标注

Face Detection and Annotation Based on a Family Digital Photo Album

针对家庭数字相册提出一种对数码照片进行人脸检测,并给检测出的人脸进行人名标注的管理方法。首先利用Adaboost方法对家庭数字相册中的照片进行人脸检测。通过对人脸图像进行灰度归一化和几何归一化等预处理得到标准的人脸图像。利用主分量分析与核主分量分析2种方法进行人脸特征提取,并对比其试验结果,最后采用欧式距离来度量人脸相似度,最终得到人脸相似度匹配的姓名推荐列表。

This paper presents a management method to detect a face and annotate a name for the face in a family digital photo album. First, the Adaboost method is used to detect the faces of the photos in the family digital album. Standard face image is obtained through preprocessing the face image normalization and geometric normalization. The principal component analysis and the kernel principal component analysis are used for facial feature extraction, and the test results are compared. Finally, Euclidean distance is used to measure similarity of faces, and face similarity matching name recommendation list is obtained.