基于小波变换和余弦变换的Fisher脸识别

结合几种现有的人脸识别特征提取算法，先对人脸图像进行小波分解去噪；然后通过离散余弦变换对低频分量作进一步特征提取和压缩，保留人脸图像中对光照、姿态、表情变化不敏感的识别信息；接着利用PCA和LDA相结合得到最终的识别特征，最后采用欧式距离和最近邻分类器识别人脸。实验采用ORL标准人脸库验证了这种组合的有效性。

Fisher face recognition based on wavelet transform and cosine transform

This article combines several existing feature extraction algorithms for face recognition. Firstly, face images are decomposed by using wavelet transform, in which some noises have been removed from the images. Then, discrete cosine transform is used on low frequency components to get further feature extraction and compression, which is not sensitive to light, gesture or facial expression. After that, a combination of PCA and LDA is conducted to obtain final face features. Finally, Euclidean distance and the minimum distance classifier are used to perform face recognition. The simulation experiments based on ORL show a better recognition rate in this combination.