Vector Approximation Based Indexing for High-Dimensional Multimedia Databases
One of the fundamental problems, in multimedia databases domain, resides in the similarity search, i.e. the need to retrieve a small set of objects which are similar or closest to a given query object. Generally, the similarity is not measured on the multimedia objects directly, but on their traditional primitives (histograms of colors and signatures sound). These primitives appear as vectors of numeric values known as feature vectors and constitute the indexes of these objects. This way, similarity searching in multimedia database becomes a K-Nearest Neighbor (K-NN) search in a high-dimensional vector space.