Abstract: This paper proposes a face recognition approach of two-dimension class-augmented PCA. Human face features are obtained by 2DPCA method. These features are normalized. The normalized features are combined with the class information to construct the class-augmented matrix. The class-augmented matrix features are computed by 2DPCA method again. In the proposed method, both structure and class information are considered, therefore, the recognition rate is improved. Experimental results on Yale and FERET data set show that the proposed method is effective for human face recognition.

Key words: human face recognition, feature extraction, two-Dimension Principal Component Analysis（2DPCA）, Class-Augmented Principal Component Analysis（CAPCA）, two-Dimension Class-Augmented Principal Component Analysis（2DCAPCA）

摘要： 提出了一种二维类增广PCA（2DCAPCA）的人脸识别算法。用二维PCA（2DPCA）方法直接对人脸图像矩阵进行特征提取，对提取的特征进行归一化处理，将归一化处理后的特征与类别信息结合构成类增广矩阵，对类增广矩阵进行2DPCA处理，提取图像的类增广矩阵特征。由于该算法既保留了人脸图像的结构信息，又考虑了样本的类别信息，识别率有了较大的提高。通过Yale和FERET库上的实验表明，该方法对人脸识别是有效的。

关键词: 人脸识别, 特征提取, 二维主成分分析（2DPCA）, 类增广主成分分析（CAPCA）, 二维类增广主成分分析（2DCAPCA）