Back to articles
General Papers
Volume: 44 | Article ID: art00013
Image Categorization
  DOI :  10.2352/J.ImagingSci.Technol.2000.44.6.art00013  Published OnlineNovember 2000

A study was conducted to investigate whether images of natural scenes can be categorized with respect to information content and whether a relation exists with perceived foreground-background separation. In an experiment, one group of subjects carried out a ‘free categorization’ task, (subjects were free to choose similarity criteria), while another group of subjects sorted images with respect to ‘foreground-background’ separation. A diverse set of photographic pictures was used in this experiment. The created clusters were analyzed by applying nonmetric multidimensional scaling and hierarchical cluster analysis. The results indicate that subjects are able to classify images in a consistent manner. Another outcome of this study is that observers classify images into categories that reflect scene content. The most striking image feature based on scene content is the presence of people or not. The second important image feature is whether these images show landscapes or objects/buildings. The results of the ‘free categorization’ task compared to the ‘foreground-background categorization’ task show some interesting correlations. Pictures of landscapes nearly always present only background information. A portrait demonstrates almost only foreground information. Pictures of objects or buildings, on the other hand, show a combination of fore- and background information.

Subject Areas :
Views 13
Downloads 0
 articleview.views 13
 articleview.downloads 0
  Cite this article 

Ina Klein Teeselink, Frans Blommaert, Huib de Ridder, "Image Categorizationin Journal of Imaging Science and Technology,  2000,  pp 552 - 559,

 Copy citation
  Copyright statement 
Copyright © Society for Imaging Science and Technology 2000
  Login or subscribe to view the content