With the growing amount of images, videos, and music, the task to support users in exploring multimedia databases is of ever-increasing importance. That is why numerous content-based browsing approaches have been developed. They support users in searching and browsing for multimedia objects in an interactive and playful way. In terms of query performance, however, these browsing approaches are frequently limited to small-to-moderate size databases. The question of how to efficiently browse large-scale multimedia databases yet remains to be answered.
In this talk, I will provide an overview of efficient query processing techniques applicable to content-based browsing systems. To this end, I will first introduce the domain of adaptive distance-based similarity measures and show how to model image similarity in a flexible way. I will then present recent developments for efficient exploration query processing by means of similarity-based visualizations and metric indexing. Finally, I will show how to browse millions of multimedia objects in a few seconds.