Currently the Visualization Group works with researchers who are doing adaptive mesh refinement, AMR, computations and/or generating large data sets. Gunther and Oliver have each been working on one aspect of the visualization problems which arise.
Gunther has been working on generating an isosurface directly from scalar AMR data. Since there is overlapping data and data at different resolutions, most straightforward extensions to existing isosurface algorithms lead to overlaps and/or cracks. The goal of this work was to find a more general technique which is still efficient but which doesn't contain artifacts.
Gunther will be presenting a method for "stitching" the AMR grids together in such a way that a seamless isosurface can be generated efficiently. This is done by using fast, existing algorithms everywhere but at the boundaries between different resolution grids. At the boundaries, the data is "stitched" together to form "cells" for which isosurfaces can be efficiently generated. This "stitching" can be applied to a much more general class of overlapping grids and it can be used as the basis for extensions of other visualization algorithms.
Oliver has been working on generating interactive isosurfaces for the exploration of large data sets in an immersive (VR) environment. A lot of scientific research these days is performed by gathering and analyzing hideous amounts of data. No matter if this data is the result of real-world measurements or of numerical simulations, extracting information from it becomes more and more difficult as the amount of data grows. Thus, scientific visualization is becoming a major tool of research: The human visual system is unparalleled in its capacity to see patterns or detect features in vast amounts of data.
The effectiveness of scientific visualization can be further enhanced by employing virtual-reality methods: The combination of true 3D (stereographic) output, 3D input devices and interactivity enables a user to explore an unknown 3D data set with unprecedented ease.
Oliver's talk will concentrate on a data exploration system he implemented as a summer project with the Visualization Group. This tool allows a user to expand an isosurface from an arbitrary point inside a data set, and to animate sets of isosurfaces by moving the point of interest.
Snacks will be provided.
See Conundrum Talks for more information about this series.