vis.uni-stuttgart.deVisualization of Astronomical Data and Objects
Introduction
In this research project, novel algorithms are developed to visualize lifelike astronomical objects efficiently in 3D. The spatial structure of planetary nebulae is reconstructed from astronomical observations and physical boundary conditions. Image data in 2D is used to calculate a physically consistent 3D model, which is then visualized with global illumination to achieve realisitic results. The aim of this project is to present cosmological phenomena with scientific facts in a visually convincing manner to the broad public by employing dome projections with high resolution, which are commonly found in planetariums. One of the main challenges is to reach interactive visualization with multi-video projectors without time consuming production of animation sequences.
 |
 |
 |
Funding and Cooperation
This project is funded by the Deutschen Forschungsgemeinschaft (DFG) for a period of three years. The project is a cooperation with Prof. Dr. Marcus Magnor from the Computer Graphics Lab at TU-Braunschweig. The main objective of their work is the spatial reconstruction of planetary nebulae from astronomical observations. At the Visualisierungsinstitut der Universität Stuttgart (VISUS), Prof. Dr. Daniel Weiskopf and Dipl.-Inf. Marco Ament are responsible for realistic and efficient visualization of the reconstructed nebulae. The parallelization of the volumetric illumination on a cluster and graphics hardware (GPU) plays an important role to achieve real-time performance on high-resolution displays. Furthermore, visual influences from the special and the general theory of relativity are taken into account in a cooperation with the project "Visualization in Special and General Relativity" to illustrate extensive cosmological effects.
Visualization Techniques
The visualization of astronomical nebulae is based on tracing light rays (ray casting). For each pixel on the display, a ray is traced through the volume and a physically-based radiative simulation calculates the final color. This method is computationally expensive, especially for large, high-resolution displays. For this reason, modern graphics hardware accelerators and parallel clusters are necessary to achieve high performance. The acceleration of the visualization is achieved by image-space decomposition, the so-called "Sort-First" approach. Every subimage is calculated independently in parallel on the compute nodes of a distributed GPU cluster. In the last step, the partial results are gathered from the nodes and are composited to the final image. One of the main challenges is the acceleration of multiple scattering because of the high dependency between the partial images, which requires expensive data transfer between the nodes.
Publications
@article {tvcg11_dvr,
author = {Moloney, Brendan and Ament, Marco and Weiskopf, Daniel and M\"{o}ller, Torsten},
title = {Sort First Parallel Volume Rendering},
year = {2011},
journal = { IEEE Transactions on Visualization and Computer Graphics},
volume = {17},
number = {8},
pages = {1164--1177},
url = {http://doi.acm.org/10.1109/TVCG.2010.116},
doi = {10.1109/TVCG.2010.116}
}