In this talk I will highlight recent research results of the graphics
group in Ulm. The first project addresses rendering of highly complex
geometry modeled by surface patches tessellated to micro-polygons. As
the number of micro-polygons can exceed the available memory we propose
ray-reordering as well as an O(N) algorithm for building the BVH over
the patches. The second project improves on the results of Monte Carlos
simulations which inherently exhibit noise, especially in real-time
settings where convergence is not yet fully reached. The use of
edge-avoiding à-trous wavelets allows us to filter the coarsely sampled
global illumination solution in real-time while correctly preserving all
geometry and illumination edges. Augmenting the framework by improved
edge-sensitivity we can further perform highly efficient image
processing such as denoising and contrast enhancement significantly
faster than state-of-the-art. At the end, I will briefly present our
newest acquisition device: a hyperspectral light stage.