Top N optimization issues in MM databases
Henk Ernst Blok · University of Twente Research Information · 2000
In multi media (MM) DBMSs the usual way of operation in case of a MM retrieval query is to compute some ranking based on statistics and distances in feature spaces. The MM objects are then sorted by descending relevance relative to the given query. Since users are limited in their capabilities of reviewing all objects in that ranked list only a reasonable top of say N objects is returned.