This project will develop a preliminary study of Quasi-Monte Carlo methods as applied to fast isosurface extraction. This study is different from general isosurface extracting methods that work on the whole data set. The PI's methods are designed to utilize only a subset of the original large three-dimensional data set that is generated by Quasi-Monte Carlo techniques. The isosurface is generated on this subset as an approximation to the isosurface generated from the whole data set. The PI proposes to initially study the Hammersley, Halton and Hyperbolic Cross points as the Quasi-Monte Carlo points in the implementation.
Preliminary tests show that the QMC techniques enjoy a linear speedup with the number of QMC points. For large data sets, the PI usually can reduce the data size remarkably and still get a good representation of the original isosurface. The advantage of these techniques becomes more prominent when the data size gets larger. Dr. Sikorski believes that the QMC points will generally generate visually better and smoother isosurfaces. These isosurfaces will represent the overall shape of the original isosurfaces better than isosurfaces based on a regular subset of the original data. The preprocessing cost of the QMC isosurface extraction may be high. This is however a one-time process. After it is completed, the post-isosurface extraction is very efficient.
The success of the project will be measured by applying their techniques in computational fluid dynamics (CFD) as well as in combustion engineering application under the ASCI (Accelerated Strategic Computing Initiative) project. By working with ASCI scientists at the University of Utah Dr. Sikorski will investigate visualization of the flows around various container structures in the simulation of accidental fires and explosions.
