Incorporating Application-Level Knowledge
into the MPEG-2 Coding Model
University of North Carolina at Chapel Hill
Current multimedia standards (in particular video
standards) are inflexible and designed for specific
applications, often without regard for heterogenous
network environments like the Internet.
Application-level semantics about the content and
context of video information should be reflected in
how video is coded and transmitted.
Current standards are inflexible because interpretation
and management of different elements of the
video stream are implicit and intertwined. Encoded
video formats can be recast into a more
flexible model by separating the generation, management,
and transmission of different video elements. A more flexible
coding framework will allow applications to better manage
network transmission of video stream elements in a manner that
reflects application-level knowledge and requirements.
Furthermore, the coding model can be
adapted to changing network conditions. This paper
specifically addresses how reference frame management
in the MPEG standards can be recast into
a more flexible coding model. By separating the
management of reference information from the coding of
specific frames, we enable the standard to be
more effectively used when application-level knowledge
of the video content is present. A number of
interesting network strategies are enabled by the
proposed revised model.