Abstract

Partial absence of face information challenges the robustness of face recognition algorithms. In order to reduce the effect of partial information loss on face recognition, a Face Recognition Method based on partitioning Collaborative Representation (FRAPCR) is proposed in this paper. Firstly, the face image is divided into several sub-blocks. Secondly, the Collaborative Representation (CR) is used to calculate the minimum sparse representation coefficient of each sub-block and the residual between the sub-block and the corresponding samples of each class, taking the class corresponding to the minimum residual as the class to which the sub-block belongs. Thirdly, a voting mechanism is introduced to count the categories of all sub-blocks of each face image, and the category with the largest number of votes is the category to which the whole face image belongs. Through the experiments on face databases ORL, Extend Yale B, and AR by the proposed method (FRAPCR), the best partitioning way of the face image is obtained. When there is partial information missing (pixel information missing, corrosion block and occlusion) in the face image, the images in each face database is divided in its corresponding optimal partitioning way. And comparative experiments between the FRAPCR and traditional CR face recognition methods are performed. The results show that FRAPCR has high recognition rate and stable recognition effect when there is partial information missing in face images.

摘要

人脸信息的部分缺失挑战了人脸识别算法的鲁棒性。为了减少部分信息丢失对人脸识别的影响，本文提出了一种基于分区协同表示（FRAPCR）的人脸识别方法。首先，人脸图像被分成几个子块。其次，利用协同表示（CR）计算每个子块的最小稀疏表示系数以及子块与每个类对应样本之间的残差，取最小残差对应的类作为类子块属于哪个。第三，引入投票机制对每张人脸图像的所有子块的类别进行计数，得票最多的类别就是整张人脸图像所属的类别。通过对人脸数据库ORL、Extend Yale B和AR的实验，利用所提出的方法（FRAPCR），得到了人脸图像的最佳分割方式。当人脸图像中存在部分信息缺失（像素信息缺失、腐蚀块和遮挡）时，将每个人脸数据库中的图像按其对应的最优划分方式进行划分。并进行了FRAPCR与传统CR人脸识别方法的对比实验。结果表明，当人脸图像存在部分信息缺失时，FRAPCR具有较高的识别率和稳定的识别效果。得到人脸图像的最佳分割方式。当人脸图像中存在部分信息缺失（像素信息缺失、腐蚀块和遮挡）时，将每个人脸数据库中的图像按其对应的最优划分方式进行划分。并进行了FRAPCR与传统CR人脸识别方法的对比实验。结果表明，当人脸图像存在部分信息缺失时，FRAPCR具有较高的识别率和稳定的识别效果。得到人脸图像的最佳分割方式。当人脸图像中存在部分信息缺失（像素信息缺失、腐蚀块和遮挡）时，将每个人脸数据库中的图像按其对应的最优划分方式进行划分。并进行了FRAPCR与传统CR人脸识别方法的对比实验。结果表明，当人脸图像存在部分信息缺失时，FRAPCR具有较高的识别率和稳定的识别效果。并进行了FRAPCR与传统CR人脸识别方法的对比实验。结果表明，当人脸图像存在部分信息缺失时，FRAPCR具有较高的识别率和稳定的识别效果。并进行了FRAPCR与传统CR人脸识别方法的对比实验。结果表明，当人脸图像存在部分信息缺失时，FRAPCR具有较高的识别率和稳定的识别效果。