BI & TRI DIMENSIONAL SCENE DESCRIPTION AND COMPOSITION IN THE MPEG-4 STANDARD

MPEG-4 is a new ISO/IEC standard being developed by MPEG (Moving Picture Experts Group). The standard is to be released in November 1998 and version 1 will be an International Standard in January 1999 The MPEG-4 standard addresses the new demands that arise in a world in which more and more audio-visual material is exchanged in digital form MPEG-4 addresses the coding of objects of various types. Not only traditional video and audio frames, but also natural video and audio objects as well as textures, text, 2- and 3-dimensional graphic primitives, and synthetic music and sound effects. Using MPEG-4 to reconstruct an audio-visual scene at a terminal, it is hence no longer sufficient to encode the raw audio-visual data and transmit it, as MPEG-2 does m order to synchronize video and audio. In MPEG-4, all objects are multiplexed together at the encoder and transported to the terminal Once de-multiplexed, these objects are composed at the terminal to construct and present to the end user a meaningful audio-visual scene. The placement of these elementary audio-visual objects in space and time is described in the scene description of a scene. While the action of putting these objects together in the same representation space is the composition of audio-visual objects. My research was concerned with the scene description and composition of the audio-visual objects that are defined in an audio-visual scene Scene descriptions are coded independently irom sticams related to primitive audio-visual objects. The set of parameters belonging to the scene description are differentiated from the parameters that are used to improve the coding efficiency of an object. While the independent coding of different objects may achieve a higher compression rate, it also brings the ability to manipulate content at the terminal. This allows the modification of the scene description parameters without having to decode the primitive audio-visual objects themselves. This approach allows the development of a syntax that describes the spatio-temporal relationships of audio-visual scene objects. The behaviours of objects and their response to user inputs can thus also be represented in the scene description, allowing richer audio-visual content to be delivered as an MPEG-4 stream.