Content-Based Image Retrieval Based on Local Affinely Invariant Regions

This contribution develops a new technique for content-based image retrieval. Where most existing image retrieval systems mainly focus on color and color distribution or texture, we classify the images based on local invariants. These features represent the image in a very compact way and allow fast comparison and feature matching with images in the database. Using local features makes the system robust to occlusions and changes in the background. Using invariants makes it robust to changes in viewpoint and illumination. Here, \similarity" is given a more narrow interpretation than usual in the database retrieval literature, with two images being similar if they represent the same object or scene. Finding such additional images is the subject of quite a few queries. To be able to deal with large changes in viewpoint, a method to automatically extract local, aanely invariant regions has been developed. As shown by the rst experimental results on a database of 100 images, this results in an overall system with very good query results.