Content area
Abstract
Existing methods of image compression are approaching the limits of the bit-rate fidelity compromise. To achieve further compression coding effort needs to be focussed on objects and regions of subjective interest within the image scene whilst relaxing efforts in other areas.
The newly standardised MPEG IV permits the description of shape which opens the way for object based image coding. The introduction of objectivity to image coding is hampered by difficulties in reliable segmentation. The use of a priori knowledge about the scene contents and more general segmentation and modelling techniques simplifies an otherwise ill defined problem.
The work to follow describes and constructs a number of different types of model-based coder and addresses the implementational issues of model-based remote conference video in a real network type environment.