Narrow Results

Annotating photographs with broad semantic labels can be useful in both image processing and content-based image retrieval. We show here how low-level features can be related to semantic photo categories, such as indoor, outdoor and close-up, using decision forests consisting of trees constructed according to CART methodology. We also show how the results can be improved by introducing a rejection option in the classification process. Experimental results on a test set of 4,500 photographs are reported and discussed.

There are millions of digital images available and their use has been increasing. Image users are very diverse, from professional users to users searching for fun. An image can be seen and described in a different way by different users. What the user sees depends on his perception and cognitive processes. The purpose of this paper is to examine how users describe images. Students of the University of Ljubljana are surveyed on a random sample of images taken from Digital Library of Slovenia and Flickr in order to assign tags to each image. The tags are grouped into categories of image attributes such as objects, place, time, etc. and compared to tags assigned by Slovene library professionals and Flickr visitors. Finally, the obtained data are investigated to ascertain whether differences in tagging practices exist among students, library professionals and Flickr users.