Research
Research Topics
Interactive Global IlluminationGlobal illumination computations are often a pre-process, resulting in a solution that can be rendered interactively as walkthroughs. For dynamic lighting conditions however, a pre-computation is often not possible or too memory consuming. Therefore plausible approximations of global illumination effects have to be computed instead. In this field of research I have developed methods for the interactive rendering of translucent materials and indirect illumination.
In a more recent project (Scalable Global Illumination) I developed a novel formulation of the rendering equation which allows for global illumination without explicit computation of visibility. Only cheap local operations are required for computation. The benefits are easy mapping to stream processor architectures and GPUs and very good performance.
Parallel Computing, GPGPU
Exploiting the power of parallel architectures is, and will be, important for photorealistic rendering, simulations, and numerical methods in general. This requires the development of dedicated algorithms, methods, and tools to facilitate the programming of such architectures. We develope tools and extensions to existing programming languages to provide higher scalability and ease of programming, and we also use GPUs for general purpose computations in numerous projects.
Interactive Photorealistic Terrain Rendering and Procedural Models
Terrain rendering is important for applications such as flight or driving simulators, geographical information systems, landscape architecture and computer games. Photorealistic renderings at interactive frame rates are desirable and necessary for most of the applications. The terrain elevation and surface data can either be acquired from the real-world, e.g. satellite images, or generated procedurally. An important research aspect is the complementing synthesis of real-world and procedural data when the real-world data cannot be acquired at the required accuracy or cannot be stored due to its amount.
Point Based Rendering
Interactive rendering using point primitives instead of traditional triangular representations has been one focus of research. Based on previous rendering algorithms, new approaches were developed for the maximum utilization of graphics hardware and improved image quality.
Sound and Perception
Recently I started working on new things. I worked on sound rendering (with Nicolas Tsingos and Sylvain Lefebvre) to approximate first-order scattering in complex environments using GPUs. In another project I worked on rendering and experiments considering contrast and spatial masking in human perception.
Current Research Projects
- Real-time rendering for next-gen video games, funded by Crytek GmbH
- Coupled Simulation of Light and Sound Propagation in Complex Scenes, within the Cluster of Excellence on Simulation Technology
- Context-Synthesis from Simulation and Models ("Synthese von Kontextinformation durch Kombination multimodaler Sensordaten mittels prozeduraler und computergraphischer Modelle ") within the DFG-funded Collaborative Research Centre Nexus (SFB 627)
- Visualization of Light Transport, funded by the Landesstiftung Baden-Württemberg
Current PhD students
- Thomas Engelhardt (global illumination, real-time rendering, SFB 627), 2008-
- Gregor Mueckl (light and sound propagation, project within the Cluster of Excellence Simulation Technology, 2009-
- Jan Novak (global illumination, BSSRDFs), 2009-
- Gabor Liktor (research cooperation funded by Crytek on real-time rendering), 2009-
- Anton Kaplanyan (external cooperation with Crytek), 2009-
- Tim Reiner (real-time rendering and modeling), 2010-
- Thorsten Schmidt (global illumination, real-time rendering), 2010-
Older Research Projects
Scalable Global IlluminationI received a post-doctoral research fellowship of the Marie Curie Intra-European Fellowship (2006/07). I developed a scalable and flexible global illumination system which adapts to accuracy of the global illumination solution to the available computation time.
VisProMo
In the scope of the VisProMo research project (2005/06) I worked on the view-dependent evaluation of procedural models for realistic natural objects at run-time on modern, programmable graphics hardware. The main foci of my work were procedural models for realistic terrain elevation and surface texturing.
3D Image Analysis and Synthesis
I received a research fellowship of the Graduate Research Center "3D Image Analysis and Synthesis" of the University of Erlangen (2002/03). In this interdisciplinary research program I have worked on methods for realistic, interactive 3D image synthesis.
Model-Based Analysis and Visualization of Complex Scenes and Sensor Data
Within the Special Research Sector 603 "Model-Based Analysis and Synthesis of Complex Scenes and Sensor Data" (2003-05) of the German Research Foundation, I have developed methods suitable for interactive visualization of neuronal pathways.
BMW Car Driving Simulator
Within the scope of a master thesis (Student: Franz Lankes) that I supervised the BMW car driving simulator has been extended with several new rendering techniques to improve the quality of the rendering process, e.g. with advanced bump mapping techniques, an analytic model for sky and sun light and motion blur. An exploratory focus is how various rendering parameters affect the perception of speed for a subject in the simulator.