基于DCT和2D—PCA的人脸识别算法研究

提出了一种对角DCT和模块2DPCA相结合的人脸识别方法．该算法首先将人脸图像转换成对角图像，提取人脸的行、列与结构信息以求解最优识别向量．然后利用DCT压缩以去掉人眼不敏感的中频分量与高频分量，再由IDCT重建人脸图像，这样有限降低了所需特征的维数，减少了计算量．然后通过模块2DPCA进行特征提取得到人脸识别特征，最后运用最近邻分类器完成人脸的识别．基于ORL及Yale人脸数据库的实验结果证明了该算法的有效性与稳健性．

Face recognition system based on DCT and modular 2D-PCA

A face recognition method based on the combination of discrete cosine transform （DCT） and modular two-dimensional principal component analysis （2D-PCA） was put forward. Firstly a face image was turned into a diagonal image which contains the row, column and structure information to the solve optimal recognition vector. Secondly the image was processed by DCT and IDCT to reduce the character dimensions effectively. Thirdly, the image was processed by 2DPCA to obtain the face recognition characters. Finally, the nearest neighbor （NN） classifier was selected to perform face recognition. The experimental results of Yale face database which are based on ORL show that this method is efficient in the face recognition.