Computer simulations and even experimental devices generate more than a single variable. These variables typically exhibit complex interdependencies which must be thoroughly understood to improve the simulation codes and understand the complex systems being studied. Unfortunately, cluttering prevents this from being achieved in current visualization systems. Users are thus left with generating several different representations and mentally merging them together, a nearly impossible feat. The goal of this project is to provide new methods that will let scientists browse through their terabytes of data searching for interesting patterns or irregularities. A secondary component of this project is to illustrate the utility of the new technique to the rest of the computer graphics industry, including efficient volume rendering for objects such as fire, clouds and water. These solutions will bring computational scientists one step closer to a seamless environment for studying both multiple simulations and complex interactions within a simulation.
Technically, the project will investigate new methods for representing complex vector fields and multi-dimensional scalar fields defined over a three-dimensional mesh topology. A particular emphasis will be on extending a direct volume rendering technique known as "splatting" to provide integrated representations of multiple data variables. (The name comes from a step in the method that constructs a footprint or "splat" used to project 3D basis functions onto surface textures.) In particular, this research will extend the state-of-the-art in direct volume rendering by developing techniques to depict non-homogenous or isotropic volumes. These volume textures will be developed along with efficient rendering techniques to allow them to reinforce or segregate different regions of a volume, control representations of time-varying data, and apply the representations to virtual reality settings.
