MB/sup +/-tree: a new index structure for multimedia databases

We describe a new indexing structure named the MB/sup +/-tree which has been designed to assist content-based retrieval in image and video databases. The types of queries considered are the nearest-neighbor query based on a weighted Euclidean-distance similarity measure and the range query, both in a multidimensional space. The indexing structure can support queries which retrieve a large number of approximate matches as well as those which retrieve a small number of close matches. The MB/sup +/-tree is based on linear order in the multidimensional space and has several features of the standard B/sup +/-tree.