Computer Graphics Research Group

Department of Computer Science and Engineering
Chalmers University of Technology
Department of Computer Science and Engineering
SE-412 96 Göteborg, Sweden.
Tel: +46 31 772 1000. Fax: +46 31 772 3663

Contact: Ulf Assarsson

Members:


Ulf Assarsson
Professor
Head

Erik Sintorn
Assistant Professor
NVIDIA Fellowship

Viktor Kämpe
PhD
 

Dan Dolonius
PhD Student
 

Sverker Rasmuson
PhD Student
 
Previous members:

Pierre Moreau
M.Sc. Student
 

Markus Billeter
PhD
 

Ola Olsson
PhD
 

Research Focus:

Current projects:
  • Compression with Sparse Voxel DAGs.
  • Free Viewpoint Video... (more to come)...
Older projects:
  • GPGPU - Modern graphics cards can be viewed as Shared memory multiprocessor systems with hundreds of small computational processing units working with thousands of threads. The fact that the speed of GPUs has doubled in speed about each year, which is a significantly faster pace than for CPUs, and that GPUs often are cheaper in terms of flops, makes them highly interesting to utilize for industrial computations. The memory system works differently than in a standard computer and coherent memory accesses is imperative for achieving maximum performance. The research group is specializing in parallelizing algorithms for one or more GPUs.
    Link
  • Hard and Soft Shadows - While real-time hard shadows typically is considered to be very fast to compute and therefore a nearly solved problem, the two popular solutions shadow maps and shadow volumes still suffer from aliasing/biasing and high fillrate demands respectively. The goal is to achieve algorithms allowing alias-free shadows from hundreds of light source in real-time. Futhermore, current solutions for artifact-free soft shadows in real-time are still highly computationally demanding, which requires further research. In addition, we have focused on how to compute shadows in hair and fur in real-time, which is imperative for a realistic appearance.
  • GPU Ray Tracing - While scene-hierarchies are very well explored, ray hierarchies are nearly unexplored. Ray hierarchies can be orders of magnitudes faster to construct, making them highly suitable for dynamic scenes. In addition, they scale sub-linearly when the image resolution is increased. Quadrupling the resolution only doubles the runtime, which is not the case for classic scene-hierarchy based ray tracing. This feature becomes increasingly pleasant as movies and games move to higher resolutions.


  • Publications

Projects

  • Erik Sintorn - NVIDIA Fellowship 2009. VR Grant: Alias Free Hard Shadows, Soft shadows, shadows in hair, GPGPU-algorithms and GPU ray tracing
  • Ola Olsson - Swedish Foundation for Strategic Research, EU/Vinnova: GPGPU algorithms, GPU ray tracing, real-time lighting
  • Markus Billeter - Swedish Foundation for Strategic Research: GPGPU algorithms, Light Scattering
  • Viktor Kämpe - Swedish Foundation for Strategic Research: GPGPU algorithms and GPU primitives