Scalable Image Retrieval with Optimal Configuration for P2P Network Database

A peer-to-peer content-based image retrieval system (P2P CBIR) is proposed in this paper. To co-operate with the highly regular P2P network and peer operations, the retrieval unit in one peer is designed to perform multi-instance image query with heterogeneous features and transmit relevant images to its source peer. The peer internal architecture is designed to perform the P2P image retrieval in a regular and scalable approach. The query peer can present the best retrieval results at any time, which is a compromise between the Time-To-Live (TTL) of the query message and the recall rate. Experiments show that the query efficiency (recall-rate/query-scope) of the scalable retrieval approach is better than previous methods, i.e., firework query model and bread first search. Furthermore, an optimal system configuration method is proposed to provide the highest recall rate for a certain number of on-line users. Simulations demonstrate that recall rates can be improved to 1.5 to 2.5 times larger while the retrieval processing time is reduced to 50% of the original, under the same number of on-line users.