基于压缩感知和图像分块的遮挡人脸识别

针对基于压缩感知的SRC算法对遮挡人脸识别效果不够理想的问题, 提出一种先将图像分块再进行识别的方法。将遮挡分散在尽可能少的分块中以降低遮挡对人脸识别的不利影响, 从而提高识别率。在AR人脸数据库上的实验结果表明, 使用该方法的遮挡人脸识别率可超过80%, 显著高于基本SRC算法40%~50%的识别率。  

Occluded Face Recognition Based on Compressed Sensing and Image Blocking

A method is proposed based on image blocking for the poor results of occluded face recognition using SRC algorithm based on compressed sensing.It performs recognition after blocking image.It disperses the occlusion in the less blocks as possible in order to reduce the adverse effect of occlusion on face recognition.Results of the experiments based on AR face database show that, occluded face recognition rate with this method can reach more than 80%, which is significantly higher than the rate of 40%~50% with basic SRC algorithm.