基于小波变换的二维独立元在人脸识别中应用

针对二维主元分析（Two-Dimensional Principal Component Analysis，2DPCA）和独立元分析（Independent Component Analysis，ICA）的特点，给出了二维独立元分析（Two-Dimensional Independent Component Analysis，2DICA）的概念。在2DICA算法的基础上，提出了基于小波变换（Wavelet-Transform，W1r）的2DICA（Wavelet-Transform and Two-Dimensional Independent Component Analysis，WT-2DICA）人睑识别算法。首先，利用小波变换将原始图像的高频分量和低频分量进行不同程度的分离，并忽略高频分量，获得原始图像的基本特征；然后，利用2DICA算法术得投影特征；最后依据曩近郐法则完成人脸识别，基于ORL（Olivetti Research Laboratory）与Yale人脸数据库的实验结果表明，WT-2DICA算法正确识别率高于2DPCA算法与2DICA算法，是一种有效的人脸识别方法。

2DICA Based on Wavelet Transformation and Applications in Face Recognition

Combined with the traits of Two-Dimensional Principal Component Analysis (2DPCA) and Independent Component Analysis (ICA), the concept of Two-Dimensional Independent Component Analysis (2DICA) is presented in this paper. Firstly, dimension reduction is done to the preprocessed face images by way of 2DPCA, and the whitened matrix is obtained. Then, the independent components of face images are acquired by way of ICA. Finally, independent basis subspace is constructed by independent basis of face images, thereby face recognition can be fulfilled according to the projected features of testing samples on the independent basis subspace. In this paper,a novel method for Two-Dimensional Independent Component Analysis based on Wavelet-Transform (WT-2DICA) in face recognition is presented. Firstly, original images are decomposed into high-frequency and low-frequency components with the help of Wavelet Transform (WT), and high-frequency components are ignored, so the prime features of original images can be attained. Secondly, the projected features are solved by 2DICA. Finally, face recognition can be realized according to the nearest neighbour rule. Experimental results on ORL (Olivetti Research Laboratory) and Yale face database show that correct recognition rate by WT-2DICA is higher than that by 2DPCA and 2DICA respectively, and the method in this paper valid in face recognition.