Efficient High-Quality Rendering Using Regular Sample-based Representations

To create a convincing image on a computer is very difficult. The complexity of the various physical pheno­mena, but also the amount of details that is necessary for a realistic appearance represent significant chal­lenges we need to overcome to enable a truly realistic image synthesis. 

Our work aimed at high-quality rendering of complex scenes in order to narrow the gap between the virtual and real world. We focussed on the question of how to create and represent data, how to structure computa­tions, and how to control their accuracy. The idea of our project was to exploit the high benefits that arise from regularity, which often translates into simpler algo­rithms and strong performance benefits. We applied this strategy to several different physical phenomena, such as lens blur, environmental ligh­ting, global illumi­nation, or reflec­tions. We made use of this aspect also in the context of content creation, where we facilitate the handling of large data sets. 

Our work led to several disseminations in high-ranked journals (CGF, TVCGJ and conferences (SIGGRAPH, Eurographics, and morel. Further, our results have an important practical impact. While a solution similar to our bent-cone approach was developed independently for the triple AAA title Crysis 2, our voxel-based global-illumination approach is actually one of the main features of the Unreal Engine 4. 

Our contributions helped creating a bridge between the virtual and real world in the context of the Intel VCI and led to an important leap towards more effective rende­ring. 

Project Team

Principal Investigator

Prof. Dr. Elmar Eisemann