结合投影算子与小波变换的人脸识别方法

Fusion of projection operator and wavelet transform for face recognition

基于投影算子具有子空间分解和子空间最佳逼近的性质和小波变换与投影空间的关系,提出了一种结合投影算子和小波变换的人脸识别方法。首先对人脸图像进行小波变换,接着在变换后的尺度空间和小波空间构造投影算子的零空间,进而构造出投影算子,最后在投影算子产生的某一子空间使用最近邻分类器进行人脸识别。通过在ORL和YALE两个人脸数据库进行与传统方法(主成分分析(PCA),线性鉴别分析(LDA),局部保距投影(LPP),局部线性嵌入(LLE))的对比实验,发现所提方法在识别率上要高于传统方法。

Projection operator can be used to decompose the space with the best approximation and can be in connection with the wavelet transform. A fusion of projection operator and wavelet transform for face recognition was proposed. Firstly wavelet transform was used to decompose the face image, secondly null space of projection operator was constructed in the scale space and wavelet spaee in order to get a projection operator, thirdly the nearest neighbor classifier was used to recognize the face in the projection subspace. The experimental results on ORL and YALE database show that the proposed approach is superior to the methods of Principal Component Analysis （PCA）, Linear Discriminant Analysis （ LDA）, Locality Preserving Projection （LPP） and Local Linear Embedding （LLE） in terms of recognition accuracy.