中文摘要：

针对含光照、表情、姿态、遮挡等误差或被噪声污染的人脸图像的识别问题，本文提出一种基于Gabor低秩恢复稀疏表示分类的人脸图像识别方法。该方法首先用低秩矩阵恢复算法求得训练样本图像对应的误差图像；然后，对每一个训练样本图像及其对应的误差图像进行Gabor变换，得到相应的Gabor特征向量，并将这些Gabor特征向量组成一个Gabor特征字典；进而，计算测试样本图像Gabor特征向量在该Gabor特征字典下的稀疏表示系数，并用该稀疏表示系数和Gabor特征字典，对测试样本图像的Gabor特征向量进行类关联重构，同时计算相应的类关联重构误差。最后，根据测试样本图像Gabor特征向量的类关联重构误差，实现对测试样本图像的分类识别。在CMU PIE、Extend-ed Yale B和AR数据库上的实验结果表明，本文提出的人脸图像识别方法具有较高的识别率和较强的抗干扰能力。

英文摘要：

To recognize the face images containing errors of illumination,expression,pose,occlusion,or contaminated by noise,we propose a face image recognition method via Gabor low-rank recovery sparse representation-based classification .In this method,we firstly obtain the error images of the training images using the low-rank matrix recovery algorithm,and then calculate the Gabor feature vectors of the training images and the corresponding error images via the Gabor transform algorithm .With these Gabor feature vectors,we constitute a Gabor feature dictionary .Based on the Gabor feature dictionary,we calculate the sparse representa-tion coefficients of Gabor feature vector of the given test image .For each class,we use the sparse representation coefficients associ-ated with the class and the Gabor feature dictionary to reconstruct the Gabor feature vector of the given test image .And then we cal-culate the reconstruction error between the Gabor feature vector and its approximation associated with the class .Based on the recon-struction errors associated with different class,we can accurately classify the given test image .Experimental results on CMU PIE, Extend Yale B and AR databases show that the proposed face image recognition method has a higher recognition rate and greater noise immunity .