Prefetching for content-based image retrieval

Currently, research on content-based image retrieval (CBIR) is focused on the improvement of retrieval performance using relevance feedback techniques. However, for the remote user, retrieval latency is also a critical metric of satisfaction. In this paper, we show how to integrate the prefetching mechanism into the CBIR system in order to reduce retrieval latency. In addition, we study the tradeoff between retrieval performance and prefetching performance in order to jointly optimize relevance feedback and prefetching. Simulation results show that even with a simple prefetching scheme, retrieval latency can be reduced significantly.