Exploring Flickr's related tags for semantic annotation of web images

  • Authors:
  • Hongtao Xu;Xiangdong Zhou;Mei Wang;Yu Xiang;Baile Shi

  • Affiliations:
  • Fudan University, Shanghai, China;Fudan University, Shanghai, China;National University, Singapore;Fudan University, Shanghai, China;Fudan University, Shanghai, China

  • Venue:
  • Proceedings of the ACM International Conference on Image and Video Retrieval
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Exploring social media resources, such as Flickr and Wikipedia to mitigate the difficulty of semantic gap has attracted much attention from both academia and industry. In this paper, we first propose a novel approach to derive semantic correlation matrix from Flickr's related tags resource. We then develop a novel conditional random field model for Web image annotation, which integrates the keyword correlations derived from Flickr, and the textual and visual features of Web images into an unified graph model to improve the annotation performance. The experimental results on real Web image data set demonstrate the effectiveness of the proposed keyword correlation matrix and the Web image annotation approach.