Abstract: Complex circumstances such as various expression and lighting condition make face recognition challenging.Gradient information reflects the change extent of different face images.It is sensitive to the edge and insensitive to the lighting.Face recognition based on gradient information can release the influence of the complex circumstance and have robustness to the complex circumstance.In this paper，two new face recognition methods based on gradient information are proposed，which are face recognition based on gradient amplitude and face recognition based on directional gradient.The gradient information is extracted from the input images.2DPCA or 2DFLD is used to extract features from the gradient information.The samples are classified according to the similarity.Experimental results on AR and Yale-B face database show that the proposed methods can get very high recognition accuracy.

Key words: gradient information, face recognition, 2D Principal Component Analysis（2DPCA）, 2D Fisher Linear Discriminant（2DFLD）

摘要： 光照、表情等外部条件的变化是影响人脸识别效果的重要因素。梯度信息反映了图像信息变化幅度的大小，对边缘敏感，对光照不敏感。基于梯度信息的人脸识别方法能够缓解光照等变化对人脸识别的影响，具有一定的鲁棒性。提出两种基于梯度信息的人脸识别方法，即基于梯度幅值的人脸识别方法和基于方向梯度的人脸识别方法。抽取梯度信息，借助于2DPCA或2DFLD对抽取的梯度信息进行特征抽取，通过相似性进行分类。在AR和Yale-B人脸库上的实验表明所提出的两种方法均具有较好的识别效果。

关键词: 梯度信息, 人脸识别, 二维主元分析, 二维Fisher线性判别分析