使用稀疏加权平均脸及对称脸解决单样本问题

被引:12
作者
王晓辉 [1 ]
黄伟 [1 ]
秦传波 [2 ]
田联房 [2 ]
机构
[1] 韩山师范学院计算机科学与工程系
[2] 华南理工大学自动化科学与工程学院
关键词
模式识别; 人脸识别; 稀疏表示方法; 人脸单样本问题;
D O I
暂无
中图分类号
TP391.41 [];
学科分类号
摘要
在人脸识别中,传统有效的鉴别分析方法需要更多样本评估类内散度信息。由于人脸的单样本问题,导致某些经典的方法如Fisherface和Eigenface等失效,解决的方法通常是生成各种虚拟样本来扩充训练集以实施这些算法。针对这个问题,根据人脸的对称相似理论,人脸样本的相关变化信息可以从它的对称脸上提取,提出组合原始训练样本及它的虚拟平均脸、对称脸作为训练样本集,应用稀疏理论进行加权积分融合,分两步进行识别的方法,并在ORL和FERET人脸数据库做了对比实验。实验结果表明,该方法比现有一些突出效果人脸识别方法更好,并具有一定的鲁棒性。
引用
收藏
页码:1527 / 1531
页数:5
相关论文
共 9 条
[1]
Using the idea of the sparse representation to perform coarse-to-fine face recognition.[J].Yong Xu;Qi Zhu;Zizhu Fan;David Zhang;Jianxun Mi;Zhihui Lai.Information Sciences.2013,
[2]
Using the original and 'symmetrical face' training samples to perform representation based two-step face recognition [J].
Xu, Yong ;
Zhu, Xingjie ;
Li, Zhengming ;
Liu, Guanghai ;
Lu, Yuwu ;
Liu, Hong .
PATTERN RECOGNITION, 2013, 46 (04) :1151-1158
[3]
Pose invariant virtual classifiers from single training image using novel hybrid-eigenfaces.[J].Abhishek Sharma;Anamika Dubey;Pushkar Tripathi;Vinod Kumar.Neurocomputing.2010, 10
[4]
On solving the face recognition problem with one training sample per subject [J].
Wang, Jie ;
Plataniotis, K. N. ;
Lu, Juwei ;
Venetsanopoulos, A. N. .
PATTERN RECOGNITION, 2006, 39 (09) :1746-1762
[5]
Face recognition from a single image per person: A survey [J].
Tan, Xiaoyang ;
Chen, Songcan ;
Zhou, Zhi-Hua ;
Zhang, Fuyan .
PATTERN RECOGNITION, 2006, 39 (09) :1725-1745
[6]
A new face recognition method based on SVD perturbation for single example image per person.[J].Daoqiang Zhang;Songcan Chen;Zhi-Hua Zhou.Applied Mathematics and Computation.2004, 2
[7]
Enhanced (PC)2A for face recognition with one training image per person [J].
Chen, SC ;
Zhang, DQ ;
Zhou, ZH .
PATTERN RECOGNITION LETTERS, 2004, 25 (10) :1173-1181
[8]
Making FLDA applicable to face recognition with one sample per person.[J].Songcan Chen;Jun Liu;Zhi-Hua Zhou.Pattern Recognition.2003, 7
[9]
Face recognition.[J].W. Zhao;R. Chellappa;P. J. Phillips;A. Rosenfeld.ACM Computing Surveys (CSUR).2003, 4