基于中心对称梯度幅值相位模式的单样本人脸识别  

Face recognition based on center-symmetric gradient magnitude and phase pattern

作　　者：杨恢先[1] 翟云龙[1] 蔡勇勇[1] 奉俊鹏 李球球 

机构地区：[1]湘潭大学物理与光电工程学院,湖南湘潭411105

出　　处：《光电子．激光》2015年第5期969-977,共9页Journal of Optoelectronics·Laser

基　　金：湖南省自然科学基金(14JJ3077);湖南省教育厅(13C917)资助项目

摘　　要：针对传统人脸识别算法在单训练样本情况下识别效果不佳的问题,提出一种基于中心对称梯度幅值相位模式(CSGMP)的单样本人脸识别算法。首先,提取人脸图像的梯度幅值和相位信息;然后,用一种新的中心对称局部方向模式(CSLDP)算子对梯度幅值进行编码,再将梯度相位量化到8个区间进行编码,将二者融合形成人脸图像的CSGMP特征;最后,分块统计直方图特征信息,将所有块的直方图串联后作为人脸图像的特征向量,利用最近邻分类器分类识别。在YALE和AR人脸库上进行测试的结果表明,本文所提方法简单有效,对光照变化、表情变化和部分遮挡等环境下单样本人脸识别具有较好的效果。To overcome the limitations of traditional face recognition methods for single sample,a novel method of face recognition based on center-symmetric gradient magnitude and phase pattern （CSGMP） is proposed. Firstly,gradient magnitude maps and phase maps of a face image are calculated. Secondly, a new operator named center-syrmnetric local directional pattern （CSLDP） is proposed to encode the gra- dient magnitude, and gradient phase is quantized into eight regions, then the proposed CSGMP is the combination of the binary codes of phase and CSLDP of magnitude. Finally, CSGMP feature maps are di- vided into several blocks, and the concatenated histogram calculated over all blocks is utilized as the fea- ture descriptor of face recognition. The recognition is performed by using the nearest neighbor classifier. Experimental results on YALE and AR face databases validate that the CSGMP algorithm is an out- standing method for single sample face recognition under different illumination conditions, different facial expression conditions and partial occlusion conditions.

关 键 词：人脸识别 单样本 中心对称梯度幅值相位模式(CSGMP) 中心对称局部方向模式(CS-LBP) 最近邻分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]