摘要： 提出一种融合加权对称图像的二维FDA人脸识别算法。将人脸图像分解为奇偶对称脸，并利用加权因子将奇偶对称脸重构新的人脸样本，通过二维FDA算法求解新样本图像的最优特征子空间进行人脸分类。有效融合二维FDA算法的优点，并利用人脸对称性的特征，同时进一步分析加权因子对人脸识别效果的影响，通过选取最优加权因子最大地提高识别率。在人脸图像库ORL中进行的实验结果表明，该算法有效并能获得较高的识别率。

关键词: 人脸识别, 奇偶对称性, 加权因子, 二维FDA, 图像处理

Abstract: A new method based on two-dimensional FDA by combining with weighted-symmetry face is proposed for face recognition. This new algorithm introduces the even symmetry samples and the odd symmetry samples of face images. The even symmetry samples and the odd symmetry samples are used to form new face samples by a weighted actor. The optimal feature projection space of the new face samples is calculated through two-dimensional FDA, algorithm to classify the extracted features. This new method uses effectively not only the advantages of two-dimensional FDA but also the symmetrical properties of facial images to achieve better recognition rate. The influence of different weighted factor make is analyzed in this paper. Choosing the optimal weighted factor achieves the best performance. Experimental results on ORL database show the efficiency of the new method.

Key words: face recognition, odd-even symmetry, weighted factor, two-dimensional FDA, image processing

中图分类号: 

• TP18