Exploiting frame-to-frame coherence in a virtual reality system

A uniform frame rate can be achieved by selecting a level of detail for each visible object such that all the objects can be rendered within the given frame time. With many visible objects a low rendering quality must be chosen if every frame is rendered from scratch. This paper presents a load-adaptive rendering algorithm which exploits frame to frame coherence and re-uses most of the image data generated during previous frames thus decreasing the number of polygons actually rendered by an order of magnitude. Complex distant objects are replaced by polygons with an image of the respective object mapped onto them. Dynamic updates of the images before frame rendering lets these "impostors" closely resemble the original objects. The approach makes efficient use of texture memory and can advantageously be incorporated into the uniform frame rate algorithm.