中文摘要：

针对鲁棒稀疏编码算法（Robust Sparse Coding,RSC）在姿态偏转、遮挡等环境下,特征维数高、识别率较低等问题,结合对该环境下仍具有良好鲁棒性的Gabor特征,提出了一种基于Gabor特征提取的改进的鲁棒稀疏编码算法（Gabor RobustSparse Coding,GRSC）.首先对人脸图像进行分块处理;然后作多方向和多尺度的Gabor特征提取并构造字典;接着用PCA特征脸法去除相关性和降低维数;最后用加权迭代稀疏编码算法求解得到的最优稀疏系数进行判别归类.在ORL和AR数据库上验证该算法的性能,结果表明在AR数据库上识别率高达98.9％,在ORL人脸库上具有显著的优势,同时有效缩短了识别时间,是一种比较实用的人脸识别方法.

英文摘要：

To solve the problem of robust sparse coding which has high dimension and low recognition of extracted feature in pose rotate and occlu- sion, an improved RSC face recognition method is proposed based on Gabor feature extraction that is still very robust on that condition. Firstly, the presented method blocks face image by extracting the image with multi-orientation and multi-scale Gabor feature and structuring Gabor dictionary. Then, it removes correlation characteristics and reduces the dimensions by principal component analysis. Finally, it realizes recognition by proposing an effective iterative weighted sparse coding algorithm for optimal sparse coefficient. Verifying the performance of the algorithm on AR and ORL data databases, and it shows that the face recognition rate is reached 98.92% on the AR and the results have significant advantages on ORL face database. At the same time, it can effectively shorten the recognition time. Therefore, it is a feasible face recognition method in practical applications.