Individual Study Option: Scalable Multimedia Database Indexing

Most image or video search engines operate similarity search by extracting and storing feature vectors from the multimedia objects. Thus, the similarity search is transformed into a search of points in the feature space that are close to a given query point in the high dimensional feature space. Multidimensional indexing structures are supposed to cut this process short and quickly return the most similar entries without looking through the entire database. The objective of this individual study option is to research multidimensional indexing structures that are suitable for large-scale multimedia databases. This ISO report first states the major problem in querying high dimensional data (namely, the curse of dimensionality), then investigates a number of approaches to high dimensional database indexing. Conclusions are drawn with a view to practical implementations of algorithms in a follow-up project.