Authors
Janez Brank,
Dunja Mladenic,
Marko Grobelnik,
Publication date
Publisher
Total citations
Cited by
Description
Although the Semantic Web and related technologies are usually focused on textual or structured data, it is also important to consider techniques for handling other sources of data. Of particular interest are multimedia data such as images and sound and video clips. In this report we present an overview of techniques that can be used to represent images as a first step towards further processing, for example for clustering, categorization, or retrieval. We describe a software module that can be used to automatically extract several types of representations from images or sets of images. The representations currently supported include histograms, autocorrelograms, and banded autocorrelograms. As an example of use of these representations, we present results of experiments in which these representations were used as a basis for image categorization.