AUTOMATIC FEATURE EXTRACTION TO AN MPEG-7 CONTENT MODEL

摘要

Searching, filtering and browsing large digital multimedia archives is a laborious task. Finding specific information from traditional retrieval methods is inadequate as the rich detail within the multimedia content that best describes it is ignored. With MPEG-7 this need has been addressed by providing a way of describing content from a multifaceted perspective of both the semantic and structural composition of any multimedia document. In this paper we describe indexing methods for the structural components of the media to describe scenes, shots and objects. Semantics are added to these foundations by labelling and defining relationships both spatial and temporal to create a content model that is comprehensive and granular in description.