Interactive Exploration of Image Collections

Abstract

With image collections, both private and commercial, ever growing, efficient and effective tools for managing these repositories are becoming increasingly important. Content-based approaches, which are based on the principle of image feature extraction and similarity calulation based on these features, seem necessary as most images are unannotated. However, typical content-based retrieval systems have shown only limited usefulness. In this paper,we present interactive image database browsing systems as an alternative to retrieval approaches. Exploiting content-based concepts, image collections can be visualised so that visually similar images are located close in the visualisation space. Once an image collection has been displayed, the user is given the possibility of interactively exploring it further through various browsing operations. After introducing basic approaches to visualising and browsing image collections, we then focus on some of the systems that we have developed in our lab for this purpose. In particular, we look in detail at the Hue Sphere Image Browser and Honeycomb Image Browser systems, both of which provide hierarchical browsing approaches that give access to large image collections in an intuitive yet efficient manner.