中文摘要：

针对人脸识别中的遮挡和姿态偏转等问题，提出了一种基于分块LBP和鲁棒核编码（Robust Kernel Coding，RKC）的人脸识别算法，简称LBP—RKC算法。该算法首先对人脸图像进行多级分块的LBP特征提取，得到图像的每一块统计直方图特征。然后，将特征投影到核空间中，在核空间中建立一个鲁棒的回归模型来处理图像中的异常值，并利用迭代重加权算法求解该模型。最后，计算测试样本的每一块核表示重构残差并进行分类识别。实验表明，提出的LBP—RKC算法在处理遮挡、姿态偏转等人脸问题时能取得很好的识别效果，同时算法效率较高。

英文摘要：

Focused on the issue of occlusion and pose rotate in face recognition, an improved face recognition method based on the block LBP and robust kernel coding is proposed, which is named LBP-RKC. Firstly, LBP-RKC algorithm extracted the multilevel blocking LBP features of face images, and the statistical histogram features of each block will be obtained. Then, projecting the features into a kernel space, a robust regression model is used to deal with the image outliers is built in the kernel space, and it uses the iteratively reweighted algorithm to solve this model. Finally, it classifies and recognizes the test sample by calculate kernel code reconstruction residual of each block. Experiments results show that the proposed algorithm LBP-RKC has a good recognition on dealing with the face images that have occlusion and pose rotate, and the efficiency of the algorithm is higher at the same time.