Abstract: To overcome the limitations of the traditional face recognition methods under variations in posture, expression and illumination, a method of face recognition based on Histograms of Weber Oriented Gradient（HWOG） is proposed. Differential excitation operator is firstly adopted to extract the structure and texture features of an image. Then the edge features of original image are extracted by using HOG operator. HWOG feature maps are divided into several blocks, and the concatenated histogram features calculated over all blocks is used for the feature descriptor of face recognition. Finally, the recognition is performed by using the nearest neighbor classifier. Experimental results on YALE, ORL, CAS-PEAL-R1 face databases demonstrate that proposed descriptor is effective, and also robust to variations of position, expression and illumination.

Key words: face recognition, Histograms of Weber Oriented Gradient（HWOG）, differential excitation, nearest neighbor classifier

摘要： 针对传统人脸识别算法在姿态、表情和光照等变化下而引起识别效果不佳的问题，提出一种韦伯梯度方向直方图人脸识别算法（HWOG）。利用差动激励提取图像的结构和纹理信息，利用HOG算子提取原始图像的边缘特征，分块统计直方图特征信息，将所有分块的直方图串接得到人脸图像HWOG特征，用最近邻分类器进行分类。在YALE人脸库、ORL人脸库上和CAS-PEAL-R1进行实验，实验结果表明所提算法能有效提高识别率，且对光照、表情和姿态变化有较好的鲁棒性。

关键词: 人脸识别, 韦伯梯度方向直方图, 差动激励, 最近邻分类器