Relevance Feedback

Research topic: Exploration of Multimedia Databases

Large multimedia databases are common in scientific, commercial and entertainment applications. As the amount of data grows, the effort to get insight into the data grows. Relevance Feedback is a promising approach to explore comprehensive multimedia databases. Based on adaptable distance functions, our research focuses on capturing user preferences. We therefore develop several new techniques for the Earth Mover’s Distance (EMD) and for the Quadratic Form Distance (QF) that reflect the users’ perception in the search process. In addition to an effective and an efficient search process, we develop new interactive visualization techniques to make the exploration process more accessible and more intuitive. The interaction allows for exploration of large amount of data with modest effort for the user.