Robust and Efficient Global Illumination

Stunning computer-generated imagery is almost omnipresent nowadays, and movies, video games, industrial and architectural design, and many other applications are almost unthinkable without photorealistic rendering. But these renderings do not come for free, as they still require an enormous manual and computational effort. In this project we improved on both of these aspects with novel algorithms and applications.

Our novel Path-Space Regularization enabled us to compute light transport more robustly and to handle scenes which were impossible to render using previous unbiased methods (which use ray tracing and are preferred in industrial applications). With efficient caching strategies we improved the performance of ray tracing and global illumination. Our light transport manipulation built on efficient global illumination methods to provide a novel means to enable artists to selectively edit the lighting in a scene towards a specific design goal, while maintaining the physical consistency (and hence realism) of the illumination as much as possible. Manipulating lighting phenomena required a visualization of the underlying light transport, which we improved with new techniques, such as bundling and filtering, as well as an optimal display of rendered images with a novel subpixel display method.

 

Project Team

Principle Investigators

Prof. Dr.-Ing. Carsten Dachsbacher