抄録
As the amount of multimedia data increases rapidly, many techniques have been developed to efficiently manage multimedia databases. Especially, active researches have focused on fast image indexing and retrieval in image databases because of its abundant applications. This paper presents a new search algorithm for fast image retrieval, which is based on a clustering scheme and a multi-resolution data structure. The proposed algorithm consists of three steps; clustering of a database, selection of several reliable clusters, and fast search within the selected clusters. As a feature, the algorithm adopts luminance histograms with a multi-resolution data structure. By using these multi-resolution histograms, we can dramatically reduce the total computational complexity. Intensive simulation results support that the proposed scheme provides faster searching ability and near-optimal searching accuracy.
