Abstract: Aiming at the limitation of the single face feature in face recognition and the lack of discriminative information of 2DPCA, a weighted fusion algorithm of face recognition based on 2DPCA and 2DLDA is proposed, which is according to the complementary idea. Firstly, the method of DCT is used to compress and reconstruct the original face image to filter out the middle and high frequency part that the human eye is not sensitive to in the image. Next, the face feature is extracted via the way of 2DPCA. Then, the discriminative face feature of the original face image is also extracted based on 2DLDA. At last, an adaptive weight selection method is proposed to fuse two kinds of face features and realize the classification recognition. The experimental results on ORL and YALE face database show that the proposed algorithm is effective.

Key words: face recognition, discrete cosine transform, two-dimensional principal component analysis, two-dimensional linear discriminant analysis, weighted fusion

摘要： 针对单一人脸特征在人脸识别中的局限性问题和二维主成分分析人脸特征缺少判别信息的问题，利用互补思想，提出了一种改进的二维主成分分析与二维线性鉴别分析加权融合的人脸识别算法。利用离散余弦变换对原始人脸图像进行压缩并重建，以滤除图像中人眼并不敏感的中高频部分，再利用二维主成分分析方法进行人脸特征的提取；运用二维线性鉴别分析方法提取原始人脸图像中具有鉴别性的人脸特征；最后，提出一种自适应的权值选取方法，将两种人脸特征进行加权融合以实现分类识别。在ORL和Yale人脸数据库上的实验结果证明了该方法的有效性。

关键词: 人脸识别, 离散余弦变换, 二维主成分分析, 二维线性鉴别分析, 加权融合