Iterative Re-Constrained Group Sparse Face Recognition With Adaptive Weights Learning

Iterative Re-Constrained Group Sparse Face Recognition With Adaptive Weights Learning
复制标题

具有自适应权重学习的迭代重新约束组稀疏人脸识别

DOI:
10.1109/tip.2017.2681841
复制
发表时间:
2017-05
影响因子:
10.6
通讯作者:
Wang Wanliang
Wang Wanliang
中科院分区:
计算机科学1区
文献类型:
--
作者:
Zheng Jianwei;Yang Ping;Chen Shengyong;Shen Guojiang;Wang Wanliang

文献摘要

参考文献

被引文献

相似文献

在本文中,我们通过具有自适应权重学习的迭代重约束组稀疏分类器(IRGSC)来考虑稳健的人脸识别问题。具体来说,我们提出了一种组稀疏表示分类(GSRC)方法,其中协同采用加权特征和组来编码比其他基于回归的方法更多的结构信息和判别信息。此外,我们推导了一种有效的算法来优化所提出的目标函数,并从理论上证明了收敛性。 IRGSC 有几个吸引人的方面。首先,自适应学习权重可以无缝合并到 GSRC 框架中。这将数据的局部性结构和特征的有效性信息整合到 <inline-formula> <tex-math notation="LaTeX">$l_{2,p}$ </tex-math></inline-formula>-norm 正则化中,形成统一的公式。其次,由于 <inline-formula> <tex-math notation="LaTeX">$l_{2,p}$ </tex-math></inline-formula>-norm 正则化,IRGSC 对于不同大小的训练集和特征维度非常灵活。第三,导出的解被证明是一个驻点(如果 <inline-formula> <tex-math notation="LaTeX">$p \geq 1$ </tex-math></inline-formula> 则全局最优)。对代表性数据集的综合实验表明,IRGSC 是一种鲁棒的判别分类器,在处理人脸遮挡、腐败和光照变化等方面,与最先进的方法相比,显着提高了性能和效率。
In this paper, we consider the robust face recognition problem via iterative re-constrained group sparse classifier (IRGSC) with adaptive weights learning. Specifically, we propose a group sparse representation classification (GSRC) approach in which weighted features and groups are collaboratively adopted to encode more structure information and discriminative information than other regression based methods. In addition, we derive an efficient algorithm to optimize the proposed objective function, and theoretically prove the convergence. There are several appealing aspects associated with IRGSC. First, adaptively learned weights can be seamlessly incorporated into the GSRC framework. This integrates the locality structure of the data and validity information of the features into <inline-formula> <tex-math notation="LaTeX">$l_{2,p}$ </tex-math></inline-formula>-norm regularization to form a unified formulation. Second, IRGSC is very flexible to different size of training set as well as feature dimension thanks to the <inline-formula> <tex-math notation="LaTeX">$l_{2,p}$ </tex-math></inline-formula>-norm regularization. Third, the derived solution is proved to be a stationary point (globally optimal if <inline-formula> <tex-math notation="LaTeX">$p \geq 1$ </tex-math></inline-formula>). Comprehensive experiments on representative data sets demonstrate that IRGSC is a robust discriminative classifier which significantly improves the performance and efficiency compared with the state-of-the-art methods in dealing with face occlusion, corruption, and illumination changes, and so on.
DOI: 10.1109/tip.2015.2479456
发表时间: 2013-08
影响因子: 10.6
作者:
Qiang Qiu;R. Chellappa
通讯作者: Qiang Qiu;R. Chellappa
DOI: 10.1109/tnnls.2014.2377477
发表时间: 2015-02
影响因子: 10.4
作者:
Jinhui Chen;Jian Yang;Lei Luo;J. Qian;W. Xu
通讯作者: Jinhui Chen;Jian Yang;Lei Luo;J. Qian;W. Xu
DOI: 10.1109/icip.2011.6116666
发表时间: 2011-12
期刊: 2011 18th IEEE International Conference on Image Processing
影响因子: --
作者:
Yu-Wei Chao;Yi-Ren Yeh;Yu-Wen Chen;Yuh-Jye Lee;Y. Wang
通讯作者: Yu-Wei Chao;Yi-Ren Yeh;Yu-Wen Chen;Yuh-Jye Lee;Y. Wang
DOI: --
发表时间: 2005
期刊: --
影响因子: --
作者:
E. Candès;J. Romberg
通讯作者: E. Candès;J. Romberg
DOI: 10.1049/iet-cvi.2014.0114
发表时间: 2015-04
期刊: IET Comput. Vis.
影响因子: --
作者:
Dexing Zhong;Zichao Xie;Yan-Rui Li;Jiuqiang Han
通讯作者: Dexing Zhong;Zichao Xie;Yan-Rui Li;Jiuqiang Han