Tagging photos using users' vocabularies

  • Authors:

  • Xueming Qian;Xiaoxiao Liu;Chao Zheng;Youtian Du;Xingsong Hou

  • Affiliations:

  • School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an 710049, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an 710049, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an 710049, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an 710049, China;School of Electronic and Information Engineering, Xi'an Jiaotong University, Xianning Road, Xi'an 710049, China

  • Venue:
  • Neurocomputing
  • Year:
  • 2013

Quantified Score

Hi-index 0.01

Visualization

Abstract

Online social image share websites such as Flickr and Panoramio allow users to manually annotate their images with their own words, which can be used to facilitating image retrieval and other image applications. The smart-phones have made it possible for users to capture images as well as get the geographical ordinates. It is easily recognized and accepted that visually similar images captured in the same place or in the same period of time may be also relevant in contents. In this paper we propose a personalized photo tagging approach by using users' own vocabularies. It can recommend users preferred tags for their newly uploaded photos based on the history information in their social communities by modeling users' tagging habit. The fundamental idea of our approach is that we try to recommend tags to users by accumulating votes from the candidate images. The candidate images are selected in term of three factors: visual features, geographical coordinates and image taken time. Thus, the candidate images include visually similar images, images captured in the same geographical coordinates or in the same period of time. Based on these three factors, we implement seven experiments. Experimental results on a Flickr image collection of nearly 2 million images of 5607 users demonstrate the effectiveness of our approach. The experimental comparison shows that the three factors have certain effectiveness in image tagging. The image tagging approach by fusing the image taken time, GPS information, and visual features achieve satisfactory performance . The impacts of history information and the batch tagging behavior to the image tagging performances are discussed.