Network-based 3D graphics

Daniel Cohen-Or

Department of Computer Science
School of mathematical science
Tel Aviv University
Schreiber building, room 019
Tel Aviv 69978, Israel

+972-3-6408828 (voice)
+972-3-6409357 (fax)

daniel@math.tau.ac.il 

The emergence of network-based visual computing introduced new challenges to computer graphics. As ever-larger datasets are being used for remote manipulation and display, limited network bandwidth and high transmission latency have become a critical bottleneck for interactive rate remote walkthroughs. Despite the hardware and software advances on single workstation graphics in recent years, several issues regarding network-based virtual environments remain unaddressed.

 Significant research has been carried out for handling large datasets on single graphics workstations. Current solutions are based on visibility culling, levels-of-detail, and image-based techniques. All these techniques are not an inherent part of the standard graphics pipeline, but are applied to the database as a preprocess to avoid feeding redundant data to the rendering process itself. These techniques were developed under the assumption that the rendering power, and in particular the rasterization process, is the main bottleneck.

 For network-based remote walkthroughs the rendering pipeline is stretched over the network. Well-known techniques should be revised for the longer pipeline in which the graphics database (the scene) is stored in a remote server, the rendering process runs locally at the client, and the network connects them. Clearly, it is inefficient to download the entire database to the client since it may require prohibitive setup time, and the dataset might well be too large to fit in a common client.

 In my talk I will present new techniques for streaming graphical data over the network to enable web-based remote walkthrough. In particular I will focus on the streaming techniques of the (i) visibility set, (ii) synthetic video and (iii) triangular meshes.