摘要： 鉴于Gabor特征对光照、表情等变化具有鲁棒性，在寻找局部细节特征和全局轮廓特征的描述方面，提出一种基于多级局部多通道Gabor变换序列特征的人脸描述与识别方法。对人脸图像进行多级分块和对局部子块进行多方向、多分辨率Gabor小波滤波，并提取其对应不同方向、不同尺度的多个Gabor幅值域图谱(LGMM)，将各级子图像的图谱LGMM进行连接后形成多级Gabor幅值域图谱，使用径向基网络对特征进行识别。对人脸库ORL和YEL的识别实验进行对比，结果验证了该方法的有效性。

关键词: 人脸识别, 多通道Gabor, 多级Gabor, 幅值域图谱, 径向基网络

Abstract: Since Gabor feature is robust to illumination and expression variations and is successfully used in face recognition area. A multi-degree local multi-channel Gabor sequence feature for face description and identification is proposed in view of the existence of local block number problem. The proposed method decomposes the normalized face image by convolving the face image with multi-scale and multi-channel Gabor filters to extract their corresponding Local Gabor Magnitude Map(LGMM). The input face image is described by multi-scale LGMM, and features are recognized through Radical Basis Function(RBF) network. Results compared with the published results on ORL and YALE face database of changing illumination, expression and aging show the validity of the proposed method.

Key words: face recognition, muti-channel Gabor, muti-scale Gabor, magnitude map, Radical Basis Function(RBF) network

中图分类号: 

• TP18