VAICo: Visual Analysis for Image Comparison

Johanna Schmidt, Meister Eduard Gröller, Stefan Bruckner
VAICo: Visual Analysis for Image Comparison
IEEE Transactions on Visualization and Computer Graphics, 19(12):2090-2099, December 2013.
Content:

Information

Abstract

Scientists, engineers, and analysts are confronted with ever larger and more complex sets of data, whose analysis poses special challenges. In many situations it is necessary to compare two or more datasets. Hence there is a need for comparative visualization tools to help analyze differences or similarities among datasets. In this paper an approach for comparative visualization for sets of images is presented. Well-established techniques for comparing images frequently place them side-by-side. A major drawback of such approaches is that they do not scale well. Other image comparison methods encode differences in images by abstract parameters like color. In this case information about the underlying image data gets lost. This paper introduces a new method for visualizing differences and similarities in large sets of images which preserves contextual information, but also allows the detailed analysis of subtle variations. Our approach identifies local changes and applies cluster analysis techniques to embed them in a hierarchy. The results of this process are then presented in an interactive web application which allows users to rapidly explore the space of differences and drill-down on particular features. We demonstrate the flexibility of our approach by applying it to multiple distinct domains.

Additional Files and Images

Additional images and videos:
results-puzzle results-puzzle: Dataset Puzzle: Our approach identified all objects in the scene which are not present in all images or change their color.
results-satellite results-satellite: Dataset Satellite: Our approach identified the satellite image which shows damage caused by a tsunami on a coast-line in Indonesia.
video video: Demo Video - 3:50min - 27MB - DivX5
Additional files:
fast-forward fast-forward: Fast-Forward Video - 30s - 8MB - WMV
paper paper: Final version of the paper - 9MB

BibTeX

Download BibTeX-Entry
@article{schmidt-2013-vaico,
  title =      "VAICo: Visual Analysis for Image Comparison",
  author =     "Johanna Schmidt and Meister Eduard Gr{\"o}ller and Stefan
               Bruckner",
  year =       "2013",
  abstract =   "Scientists, engineers, and analysts are confronted with ever
               larger and more complex sets of data, whose analysis poses
               special challenges. In many situations it is necessary to
               compare two or more datasets. Hence there is a need for
               comparative visualization tools to help analyze differences
               or similarities among datasets. In this paper an approach
               for comparative visualization for sets of images is
               presented. Well-established techniques for comparing images
               frequently place them side-by-side. A major drawback of such
               approaches is that they do not scale well. Other image
               comparison methods encode differences in images by abstract
               parameters like color. In this case information about the
               underlying image data gets lost. This paper introduces a new
               method for visualizing differences and similarities in large
               sets of images which preserves contextual information, but
               also allows the detailed analysis of subtle variations. Our
               approach identifies local changes and applies cluster
               analysis techniques to embed them in a hierarchy. The
               results of this process are then presented in an interactive
               web application which allows users to rapidly explore the
               space of differences and drill-down on particular features.
               We demonstrate the flexibility of our approach by applying
               it to multiple distinct domains.",
  pages =      "2090--2099",
  month =      12,
  number =     "12",
  event =      "IEEE Conference on Visual Analytics Science and Technology
               (IEEE VAST)",
  journal =    "IEEE Transactions on Visualization and Computer Graphics",
  volume =     "19",
  location =   "Atlanta, GA, USA",
  keywords =   "Comparative visualization, Image set comparison,
               Focus+context visualization",
  URL =        "http://www.cg.tuwien.ac.at/research/publications/2013/schmidt-2013-vaico/",
}