Effective content representation for video
A.M. Ferman, A.M. Takalp, Rajiv Mehrotra · 2002
A universal description scheme to characterize video data and user queries is presented. We first identify content-significant components in a video sequence as defined by user requirements, and find that a representation based on people/objects, action/events, location, and time is appropriate for improving user interaction and database access. We demonstrate that currently available automatic or semi-automatic processing methods can be employed to characterize various types of content in terms of these descriptors by appropriate association of semantic meaning with syntactic features.