An efficient nearest neighbor search in high-dimensional data spaces

Similarity search in multimedia databases requires an efficient support of nearest neighbor search on a large set of high-dimensional points. A technique applied for similarity search in multimedia databases is to transform important properties of the multimedia objects into points of a high-dimensional feature space. The feature space is usually indexed using a multidimensional index structure. Then, similarity search corresponds to a range search which returns all objects within a threshold level of similarity to the query objects, and a k-nearest neighbor search that returns the k most similar objects to the query object. Initially, traditional multidimensional data structures (e.g., R-tree [1], kd-tree [5]), which were designed for indexing low-dimensional spatial data, were used for indexing high-dimensional feature vectors. However, recent research activities [10,9,8] reported