一种改进的2DPCA方法在人脸识别中的应用

摘要

为了提高二维主成份分析（2DPCA）方法在人脸识别中的识别率,提出了一种改进的2DPCA和分块图像相结合的人脸识别方法。该方法根据类内图像与该类平均图像的距离,引入加权函数,重新定义2DPCA的总体散布矩阵,并应用到分块图像中,对训练样本子图像采用改进的2DPCA方法进行特征提取,实现模式分类。在ORL标准人脸库上的实验结果表明,它可以有效地提高识别率。%To improve the face recognition rate of two dimension principal component analysis （2DPCA）, a modified method of 2DPCA based on modular images is proposed. In this paper, the weights function is introduced to define the total scatter matrix according to the distances between the within-class images and the average of the within-class images. We apply it into the modular images. The improved 2DPCA is used to extract feature in the training sample sub-images for pattern classification. Results of the experiments based on ORL face database show that it can effectively improve the rate of face recognition.