No abstract

W e present a sound framework for relevance feedback in content-based image retrieval. The modeling is based on non-parametric density estimation of relevant and non-relevant items and Bayesian inference. This theory has been successfully applied to benchmark image databases, quantitatively demonstrating its performance for target search, selective control of precision and recall an category search, and improvement of retrieval eaectiveness. The paper is illustrated with several experiments and retrieval results on real-world data.

Relevance feedback is one of the strong components of Surfimage, the INRIA content-based image retrieval system. Relevance feedback is about learning from user interaction, and is useful in tasks like query refinement and multiple queries. We present two relevance feedback techniques currently implemented in Surf image.

No abstract

No abstract