Organizing and browsing photos using different feature vectors and their evaluations

  • Authors:
  • Grant Strong;Minglun Gong

  • Affiliations:
  • Memorial University of Newfoundland, NL, Canada;Memorial University of Newfoundland, NL, Canada

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

Quantified Score

Hi-index 0.00

Visualization

Abstract

Two-dimensional similarity-based image organizing studies how to place photos within 2D virtual canvas based on their visual contents so that the users can easily locate the desired photos. As an extension to our previous work [10], several improvements are made in this paper to allow better photo browsing experiences. For example, the new approach pre-orders all the photos so that a consistent set of photos is selected for display. This solves the photo flickering problem of our previous approach, which uses K-mean algorithm to dynamically select photos. The main focus of this paper however is on the evaluation of the effectiveness of different feature vectors for 2D photo organization. A performance metric is proposed to measure how well photos with similar visual contents are grouped together on the 2D canvas. Feature vectors generated using eight different low-level feature extraction approaches are tested. The evaluation results reveal the pros and cons of different feature extraction approaches, which can be a useful guide for developing new feature vectors.