Fast exhaustive multi-resolution search algorithm based on clustering for efficient image retrieval

To find the best match for a query according to a certain similarity measure, this paper presents a fast exhaustive multi-resolution search algorithm based on image database clustering. Prior to search process, the whole image dataset is partitioned into a pre-defined number of clusters having similar feature contents. For a given query, the proposed algorithm first checks the lower bound of distances in each cluster, eliminating disqualified clusters. Next, it only examines the candidates in the surviving clusters through feature matching. To alleviate unnecessary feature-matching operations in the search procedure, the distance inequality property based on a multi-resolution data structure is employed. Simulation results show that the proposed algorithm guarantees very rapid exhaustive search. (c) 2005 Elsevier Inc. All rights reserved.