针对基于分割的遮挡人脸识别对光照不均衡敏感的问题, 提出了一种低秩分解的人脸图像光照均衡化预处理方法。首先, 基于人脸图像具有的左右对称性, 得到一组扩展的测试人脸图像; 然后, 由矢量化的扩展测试人脸图像构成矩阵, 并通过低秩矩阵分解技术提取出光照均衡化处理后的测试人脸图像用于后续的识别。基于 AR库的仿真实验结果初步验证了: 本文方法对于改进具有左右对称特征的人脸遮挡区域检测, 有效提升识别系统的性能有明显效果。

Aiming to address the problem that segmentation based face recognition is sensitive to illumination imbalance, an illumination equalization preprocessing for face image is proposed. Firstly, a set of extended test images are generated by considering the fact that the face image has symmetry. Then, the extended test face images are vectorized to concatenate a matrix, from which the illumination equalized test face image is extracted by low rank decomposition technique for the following recognition. Simulation experimental results based on AR database primarily demonstrate that the proposed method could efficiently improve the performance of detecting the bilateral symmetry occlusion of face, and thus promote recognition rate considerably.