University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program


Automatic image sorting using MPEG-7 descriptors


Prof. Dr. Kai-Uwe Barthel, FHTW Berlin
Berlin, Germany

date & place

Wednesday, 16.11.2005, 16:15 h
Room C252


1) In many areas of visual digital media many different color models and/or color spaces are used. For students most of these color spaces are difficult to understand, since they do not correspond to the human color perception. In this talk a tool will be demonstrated that can help to interactively explore different color spaces.

2) The increasing use of digital images has led to a growing problem how to organize these images in such a way that particular images can easily be retrieved or searched. Most images usually do not contain any meta- data that can be used for image retrieval or sorting. Content-Based Image Retrieval (CBIR) systems use the query-by-example principle, which suffers from a number of limitations. If no query image is available, this technique cannot be used. In addition, useful sorting of images is not possible with CBIR systems.

This talk presents an automatic image sorting technique that can help to organize large sets of images. Image retrieval can be accomplished by successively narrowing down the set of images by performing a consecutive image sorting and selection of image sub-sets. Our automatic image sorting system uses MPEG-7 visual features which are organized by a self-organizing Kohonen map.