基于类别相关近邻子空间的最大似然稀疏表示鲁棒图像识别算法

胡正平; 宋淑芬

doi:10.3724/SP.J.1004.2012.01420

基于类别相关近邻子空间的最大似然稀疏表示鲁棒图像识别算法

doi: 10.3724/SP.J.1004.2012.01420

胡正平^1,2,
宋淑芬¹

1.
燕山大学信息科学与工程学院秦皇岛 066000;
2.
河北省计算机虚拟技术与系统集成重点实验室秦皇岛 066000

计量
- 文章访问数: 1898
- HTML全文浏览量: 55
- PDF下载量: 805
- 被引次数: 0
出版历程
- 收稿日期: 2011-09-01
- 修回日期: 2012-02-22
- 刊出日期: 2012-09-20

Robust Image Recognition Algorithm of Maximum Likelihood Estimation Sparse Representation Based on Class-related Neighbors Subspace

HU Zheng-Ping^1,2,
SONG Shu-Fen¹

1.
School of Information Science and Engineering, Yanshan University, Qinhuangdao 066000;
2.
Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province, Qinhuangdao 066000

摘要

摘要: 为了构建一个快速鲁棒的图像识别算法, 提出基于类别相关近邻子空间的最大似然稀疏表示图像识别算法. 考虑到每个测试样本的不同分布特性及训练样本选择的类别代表性原则, 不再将所有训练样本作为稀疏表示的字典, 而是基于距离相近准则选择合适子空间, 从每个类别中选取自适应数量的局部近邻构成新的字典, 在减少训练样本的同时保留了稀疏表示原有的子空间结构. 然后基于最大似然稀疏表示识别方法, 将稀疏表示的保真度表示为余项的最大似然函数, 并将识别问题转化为加权的稀疏优化问题. 在公用人脸与数字识别数据库上的实验证明该算法的合理性, 提高识别速度的同时保证了识别精度和算法的鲁棒性, 特别是对于遮挡与干扰图像具有较好的适应性.
- 图像识别 /
- 稀疏表示 /
- 类别相关子空间 /
- 近邻选择 /
- 最大似然估计
Abstract: In order to construct a fast and robust image recognition algorithm, an image recognition algorithm of maximum likelihood estimation sparse representation based on class-related neighbors subspace is proposed in this paper. Considering the different distribution characteristics of each test sample and the class-representative principle of training samples' selection, instead of constructing the dictionary of sparse representation by all training samples, suitable subspace is selected and local neighbors of adaptive number that is selected from each class are used to construct the new dictionary based on distance proximity criterion. The training samples are reduced and the original subspace structure of sparse representation is kept at the same time. Then based on the recognition method of maximum likelihood sparse representation, the fidelity of sparse representation is represented by the maximum likelihood function of residuals and the recognition problem is converted to a weighted sparse optimization problem. Experiments results on public available face and handwritten digital databases verify the rationality, recognition speed, and recognition accuracy of the proposed algorithm. The algorithm is robust, especially it can work for in disturbed and occluded images.
- Image recognition /
- sparse representation /
- class-related subspace /
- neighbors' selection /
- maximum likelihood estimation

HTML全文

参考文献(1)

[1]

Wright J, Yang A Y, Ganesh A, Sastry S S, Ma Y. Robust face recognition via sparse representation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31(2): 210-227[2] Huang J Z, Huang X L, Metaxas D. Simultaneous image transformation and sparse representation recovery. In: Proceedings of the 26th IEEE Conference on Computer Vision and Image Recognition. Anchorage, United States: IEEE, 2008. 1-8[3] Wagner A, Wright J, Ganesh A, Zhou Z H, Ma Y. Towards a practical face recognition system: robust registration and illumination by sparse representation. In: Proceedings of the 2009 IEEE Computer Society Conference on Computer Vision and Image Recognition Workshops. Miami, United States: IEEE, 2009. 597-604[4] Wright J, Ma Y. Dense error correction via l1 minimization. IEEE Transactions on Information Theory, 2010, 56(7): 3540-3560[5] Yang M, Zhang L, Yang J, Zhang D. Robust sparse coding for face recognition. In: Proceedings of the 2011 IEEE Conference on Computer Vision and Image Recognition. Springs, United States: IEEE, 2011. 625-632[6] He R, Hu B G, Zheng W S, Guo Y Q. Two-stage sparse representation for robust recognition on large-scale database. In: Proceedings of the 24th AAAI Conference on Artificial Intelligence and the 22nd Innovative Applications of Artificial Intelligence Conference. Atlanta, United States: AAAI, 2010. 475-480[7] Huang J B, Yang M H. Fast sparse representation with prototypes. In: Proceedings of the 2010 IEEE Computer Society Conference on Computer Vision and Image Recognition. San Francisco, United States: IEEE, 2010. 3618-3625[8] Li C G, Guo J, Zhang H G. Local sparse representation based classification. In: Proceedings of the 2010 International Conference on Pattern Recognition. Istanbul, Turkey: ICPR, 2010. 649-652[9] Zhang N, Yang J. K nearest neighbor based local sparse representation classifier. In: Proceedings of the 2010 Chinese Conference on Pattern Recognition. Chongqing, China: CCPR, 2010. 400-404[10] Tibshirani R. Regression shrinkage and selection via the lasso: a retrospective. Journal of the Royal Statistical Society: Series B, 2011, 73(3): 273-282[11] Zhang J, Jin R, Yang Y M, Hauptmann A G. Modified logistic regression: an approximation to SVM and its applications in large-scale text categorization. In: Proceedings of the 20th International Conference on Machine Learning. Washington, United states: ICML, 2003. 888-895

施引文献

资源附件(0)

访问统计