Last modified
February 29, 2000

  Seminar Abstract
Center for Data Intensive Computing


 
 


 

Direct Volume Rendering for Scientific Visualization

Scientific data come as volumetric datasets, represented as sample points on a regular or irregular 3D raster. They are visualized on computer screens, head-mounted displays, group-immersive projection media (virtual workbench, CAVE), and large projection screens (PowerWall). Two avenues can be taken to achieve this: (1) The volumetric data are first converted into a set of polygonal iso-surfaces and subsequently rendered with polygon rendering hardware, or (2) The volumetric data are directly rendered without the intermediate conversion step. In (1) it is assumed that a set of extractable iso-surfaces exists, and further, that the infinitely thin surface the polygon mesh represents models the true object structures at reasonable fidelity. Neither is always the case, and as illustrative examples may serve: amorphous cloud-like phenomena, smoothly varying flow fields, or structures of varying depth that attenuate traversing light corresponding to the material thickness. But even if both assumptions are met, the complexity of the extracted polygonal mesh can easily overwhelm the capabilities of the polygon rendering subsystem, and a direct volume rendering may prove more efficient, especially when the object is complex or large, or when the isosurface is interactively varied and the repeated polygonal extraction overhead must be figured into the rendering cost.

In this talk, I will present the results of a recently completed quantitative and qualitative comparison of the four most popular direct volume rendering algorithms: raycasting, splatting, shear-warp, and 3D texture mapping hardware. A wide span of scenarios was covered, ranging from medical applications to scientific. One of the conclusions of this study is that high-quality volume rendering is still not feasible in real-time, at least not without using custom hardware. To address this problem, I will present a new technique, termed image-based rendering, that enables real time display by re-using computed imagery from one viewpoint for other viewpoints nearby.

Top of Page

   

 




Copyright © 1999 Brookhaven National Laboratory ALL RIGHTS RESERVED
Comments/Sugestions about this site contact: Webmaster