摘要
提出了一种基于局部奇异值对称平均的人脸识别方法。该方法首先对原始图像进行线性映射处理;接着采用局部奇异值分解提取人脸特征,并对所获得的特征作对称平均处理;最后依据最近邻决策规则进行识别。基于ORL人脸数据库的实验结果表明,该方法大大降低了原始特征空间的维数,有效地消除了图像亮度和噪声的影响,并取得了较高且稳定的正确识别率,在人脸识别中是一种有效的方法。
In this paper, a method of face recognition based on symmetry average of local singular value features is presented. First, original face image data are linearly mapped. Second, the local singular values of face image matrix are extracted and employed as the feature matrix, then the feature matrix is averaged symmetrically. Finally, nearest neighbor decision (NND) rule is used as recognition rule. Experimental results on ORL (olivetti research laboratory) database show that this method can lessen the number of original features of face image, eliminate the effects of illumination and noise of image effectively, and then get a higher correct recognition rate.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2005年第17期146-148,共3页
Computer Engineering
基金
广东省自然科学基金资助项目(032356)
关键词
人脸识别
奇异值分解
局部奇异值特征
Face recognition
Singular value decomposition: Local singular value features