Visual topic model for web image annotation

Authors:
Xianming Liu;Hongxun Yao;Rongrong Ji;Pengfei Xu;Xiaoshuai Sun;Qi Tian
Affiliations:
Harbin Institute of Technology, Harbin, Heilongjiang, P.R. China;Harbin Institute of Technology, Harbin, Heilongjiang, P.R. China;Harbin Institute of Technology, Harbin, Heilongjiang, P.R. China;Harbin Institute of Technology, Harbin, Heilongjiang, P.R. China;Harbin Institute of Technology, Harbin, Heilongjiang, P.R. China;Univeristy of Texas at San Antonio, San Antonio
Venue:
ICIMCS '10 Proceedings of the Second International Conference on Internet Multimedia Computing and Service
Year:
2010

Citing 9
Cited 0

Probabilistic latent semantic indexing

Proceedings of the 22nd annual international ACM SIGIR conference on Research and development in information retrieval
Latent dirichlet allocation

The Journal of Machine Learning Research
Matching words and pictures

The Journal of Machine Learning Research
Improving recommendation lists through topic diversification

WWW '05 Proceedings of the 14th international conference on World Wide Web
Diversifying the image retrieval results

MULTIMEDIA '06 Proceedings of the 14th annual ACM international conference on Multimedia
Real-Time Computerized Annotation of Pictures

IEEE Transactions on Pattern Analysis and Machine Intelligence
Visual diversification of image search results

Proceedings of the 18th international conference on World wide web
What is a complete set of keywords for image description & annotation on the web

MM '09 Proceedings of the 17th ACM international conference on Multimedia
Constructing Concept Lexica With Small Semantic Gaps

IEEE Transactions on Multimedia

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we focus on image semantic understanding under large scale of image set, in which traditional approaches suffer from the limitations of scalability, tag correlation and noisy items. To solve these problems, a novel Visual Topic Model framework is proposed, via unsupervised clustering techniques. The framework aims at analyzing image semantics fusing both content and context, by considering tag correlations and ambiguities. In fact, the tags highly correlated in context may vary greatly in visual content and thus represent different semantics. Furthermore, a keyword selection and image annotation algorithm is also developed and applied to Flickr database with 175,770 images. Compared with the state-of-the-art methods, credible performance provides solid support for our framework.