| The ever increasing size of scientific datasets presents a significant challenge for visualization because computers are no longer capable of displaying entire datasets at interactive rates or at rates that do not significantly slow down the visualization pipeline. Thus, there is a need for algorithms that can quickly reduce the amount of data that must visualized so that it can be drawn quickly and accurately. This dissertation presents the use of tetrahedral meshes created by edge bisection for multiresolution scientific visualization. Volume datasets, such as scalar fields or scalar fields with material interface information, defined on rectilinear grids, are embedded into a hierarchy of tetrahedral meshes. This hierarchy, which has fast, local and crack-free refinement and coarsening algorithms, is used to create a locally-adaptive approximation of the original volume based on user defined refinement criteria or error metrics. This new approximate volume representation is then used for interactive visualization and exploration via incremental view-dependent refinement and coarsening. |
