基于HOG特征和稀疏表征的鲁棒性人脸识别

HOG and Sparse Representation Based Robust Face Recognition

在图像局部遮掩或者被局部损坏的情况下，随机投影和稀疏表征的人脸识别方法可以很好的达到正确识别的目的．但是为了得到好的识别效果，这种方法对训练库中每个人的人脸图片的数目要求比较高（至少五张以上），在实际应用中很不方便，该文将HOG（方向梯度直方图）特征引入稀疏表征的人脸识别方法中能够很好的克服这一缺点。实验表明，即使在每个人的人脸图片数只有一张的情况下，该文的方法仍然能得到很好的识剐效果。

In the case of the image partially disguised or corrupted, the face recognition of random projection and sparse representation can get the purpose of correctly identified well.To get the good recognition results,the method relatively high demand for the number of every.face image in training database（at least five or more）. In practical applications it is very inconvenient.It will be well overcome this shortcoming when HOG（Histograms of Oriented Gradients） introduced in the sparse representation. Ex- periments show that even in the case of each person has only one image face, this method still can get good recognition results.