基于人脸细节方向特性的识别特征提取

充分考虑到人脸图像的整体和细节特征，进而将人脸的眉毛、眼睛、鼻子和嘴部细节部分的水平方向特性引入到特征提取环节中；将小波变换后的低频近似分量、表达上述水平特性的水平细节分量以及眉毛、眼睛、鼻子和嘴部细节区域分别进行奇异值分解，并对得到的3组奇异值进行排列组合，最终作为该图像的有效识别特征。结果表明，基于人脸细节方向特性的识别特征提取方法的识别率高于在原图上的基本奇异值分解等方法。

Feature extraction of face recognition based on the directional characteristics

This paper gives full consideration to the overall and detail features, and thus introduces the horizontal direction fea- tures to the feature extraction, such as the details of the face, eyebrows, eyes, nose and mouth. Then, it takes the singular value decomposition respectively on the areas of the low frequency approximation component and the level of the detail component which expresses above levels the characteristics after wavelet transform, and the eyebrows, eyes, nose and mouth, array and assemble the obtained three sets of singular values as the effective identification feature finally. The results show that the recognition rate is high- er than the general singular value decomposition method which based on the directional characteristics of the face recognition feature extraction.