Image Tag Completion by Noisy Matrix Recovery

It is now generally recognized that user-provided image tags are incomplete and noisy. In this study, we focus on the problem of tag completion that aims to simultaneously enrich the missing tags and remove noisy tags. The novel component of the proposed framework is a noisy matrix recovery algorithm. It assumes that the observed tags are independently sampled from an unknown tag matrix and our goal is to recover the tag matrix based on the sampled tags. We show theoretically that the proposed noisy tag matrix recovery algorithm is able to simultaneously recover the missing tags and de-emphasize the noisy tags even with a limited number of observations. In addition, a graph Laplacian based component is introduced to combine the noisy matrix recovery component with visual features. Our empirical study with multiple benchmark datasets for image tagging shows that the proposed algorithm outperforms state-of-the-art approaches in terms of both effectiveness and efficiency when handling missing and noisy tags.

[1]  Andrew J. Davison,et al.  Active Matching , 2008, ECCV.

[2]  Martin J. Wainwright,et al.  Restricted strong convexity and weighted matrix completion: Optimal bounds with noise , 2010, J. Mach. Learn. Res..

[3]  Cordelia Schmid,et al.  TagProp: Discriminative metric learning in nearest neighbor models for image auto-annotation , 2009, 2009 IEEE 12th International Conference on Computer Vision.

[4]  Kilian Q. Weinberger,et al.  Fast Image Tagging , 2013, ICML.

[5]  V. Koltchinskii,et al.  Nuclear norm penalization and optimal rates for noisy low rank matrix completion , 2010, 1011.6256.

[6]  Tat-Seng Chua,et al.  NUS-WIDE: a real-world web image database from National University of Singapore , 2009, CIVR '09.

[7]  Dong Liu,et al.  Image retagging , 2010, ACM Multimedia.

[8]  Alexandre Bernardino,et al.  Matrix Completion for Multi-label Image Classification , 2011, NIPS.

[9]  Steven C. H. Hoi,et al.  A two-view learning approach for image tag ranking , 2011, WSDM '11.

[10]  Ralf Krestel,et al.  Latent dirichlet allocation for tag recommendation , 2009, RecSys '09.

[11]  Hai Jin,et al.  Image label completion by pursuing contextual decomposability , 2012, TOMCCAP.

[12]  Lei Wu,et al.  Tag Completion for Image Retrieval , 2013, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[13]  Jian Dong,et al.  Accelerated low-rank visual recovery by random projection , 2011, CVPR 2011.

[14]  Emmanuel J. Candès,et al.  Exact Matrix Completion via Convex Optimization , 2008, Found. Comput. Math..

[15]  Rong Jin,et al.  Large-Scale Image Annotation by Efficient and Robust Kernel Metric Learning , 2013, 2013 IEEE International Conference on Computer Vision.

[16]  Jianmin Wang,et al.  Image Tag Completion via Image-Specific and Tag-Specific Linear Sparse Reconstructions , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[17]  Xiaodong Li,et al.  Dense error correction for low-rank matrices via Principal Component Pursuit , 2010, 2010 IEEE International Symposium on Information Theory.

[18]  Daniel Gatica-Perez,et al.  On image auto-annotation with latent space models , 2003, ACM Multimedia.

[19]  Hao Xu,et al.  Tag refinement by regularized LDA , 2009, ACM Multimedia.

[20]  Jieping Ye,et al.  An accelerated gradient method for trace norm minimization , 2009, ICML '09.

[21]  Inderjit S. Dhillon,et al.  Information-theoretic metric learning , 2006, ICML '07.

[22]  Emmanuel J. Candès,et al.  Matrix Completion With Noise , 2009, Proceedings of the IEEE.

[23]  David M. Blei,et al.  Probabilistic topic models , 2012, Commun. ACM.

[24]  Michael I. Jordan,et al.  Latent Dirichlet Allocation , 2001, J. Mach. Learn. Res..

[25]  Shuicheng Yan,et al.  Image tag refinement towards low-rank, content-tag prior and error sparsity , 2010, ACM Multimedia.

[26]  Ali Farhadi,et al.  Scene Discovery by Matrix Factorization , 2008, ECCV.

[27]  Dong Liu,et al.  Image Retagging Using Collaborative Tag Propagation , 2011, IEEE Transactions on Multimedia.

[28]  Heng Ji,et al.  Exploring Context and Content Links in Social Media: A Latent Space Method , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.