Tagging Image with Informative and Correlative Tags

Xiaoming Zhang,Heng Tao Shen,Zi Huang,Zhoujun Li
DOI: https://doi.org/10.1007/978-3-642-20291-9_26
2011-01-01
Abstract:Automatic tagging can automatically label images with semantic tags to significantly facilitate multimedia search and organization. Existing tagging methods often use probabilistic or co-occurring tags, which may result in ambiguity and noise. In this paper, we propose a novel automatic tagging algorithm which tags a test image with an Informative and Correlative Tag (ICTag) set. The assigned ICTag set can provide a more precise description of the image by exploring both the information capability of individual tags and the tag-to-set correlation. Measures to effectively estimate the information capability of individual tags and the correlation between a tag and the candidate tag set are designed. To reduce the computational complexity, we also introduce a heuristic method to achieve efficient automatic tagging. The experiment results confirm the efficiency and effectiveness of our proposed algorithm.
What problem does this paper attempt to address?