Hybrid representations to improve both streaming and rendering of dynamic networked virtual environment

Abstract

In this paper we present a novel combined usage of image-based and geometrical representations in dynamic networked virtual environments. Contrary to other research, the type of applications we focus on rely on a continuous streaming of representation data as the user navigates through the virtual environment. The client has no prior knowledge of the environment that needs to be rendered and the objects in the environment can be dynamic or static. Our first goal is to present the user with a sufficiently realistic view of the virtual world as soon as possible while reducing disturbing scene buildup artifacts as well as guarantee interactive frame rates. We create image-based representations that are more compact to send that geometrical models so the client will sooner be able to render a more satisfactory view than if we would only be using incrementally transmitted models. The image-based models are also used in combination with traditioal LoD techniques to speedup rendering while maintaining good display quality, which is our second goal. We also present a grouping strategy to further speedup image-based rendering. Run-time measurements will show the benefit of using our system.