Video compression implemented using rigid and non-rigid object recognition and computer graphics animation

A video compression prototype system is described that significantly reduces the required bandwidth for video transmission by using intelligent terminals. The terminals used for video compression are trained to recognize what they see. Each computer contains a three-dimensional description of objects that might be in the scene. This description includes information about the objects, such as how they move and how they are constrained. The terminal is trained to recognize certain individuals, locations, and objects that have had their features scanned into an image library. The image library contains the physical description of the objects as well as a description of how they change and move over time. Results obtained with the system are presented