Index-driven similarity search in metric spaces (Survey Article)

摘要: Similarity search is a very important operation in multimedia databases and other database applications involving complex objects, involves finding objects data set S similar to query object q, based on some similarity measure. In this article, we focus methods for that make the general assumption represented with distance metric d. Existing handling setting typically fall into one of two classes. The first directly indexes distances (distance-based indexing), while second mapping vector space (mapping-based approach). main part article dedicated survey distance-based indexing methods, but also briefly outline how occurs mapping-based methods. We present framework performing distances, algorithms common types queries operate an arbitrary "search hierarchy." These can be applied each presented, provided suitable hierarchy defined.