The Virtual Microscope

Modern material sciences try to replace costly laboratory tests by computer simulations. The focus of the virtual microscope is that, based on massive geometric data, in an acceptable time the appearance and position of visible objects in the simulation to identify and simultaneously the colors of the objects with a realistic illumination to rendering. It allows the users to recognize specific characteristics of the microstructures for different purposes such as error and pattern detection and model verification or to show effects of interactive interventions in real-time.

Real-Time Global Illumination in Computer Graphics

Global Illumination is a time consuming calculation, that cannot be implemented on the CPU in real-time. This is especially true for the simulation of the indirect light. One focus of our research is the adoption of CPU algorithms on the parallel architecture of advanced graphics hardware (GPU), so that the acceleration of the illumination method can be maximized.

Illumination for Augmented Reality

In augmented reality application the real camera picture is extended with virtual objects. Without correct illumination, these objects often appear very artificial. Through the reconstruction of the real ambient light, the virtual objects can be integrated with consistent illumination, and therefore blend into the realistic scenery. The focus in this field is the fast reconstruction of time and space varying light as additional information to the live camera picture.

Predictive Rendering

Predictive Rendering computes an exact prediction of the appearance of a three-dimensional scene, such that the rendering is practically identical to a real photograph. This requires an expensive global illumination simulation. A special challenge here is the correct simulation of complex light paths with multiple specular and diffuse reflections. One method for such simulations is Stochastic Progressive Photon Mapping (SPPM).

General Purpose GPU

Today, a GPU is a cheap, parallel processor that allows parallel computations on hundreds of threads. Beside the pure rendering, general purpose problems can be solved on a GPU. Depending on the application, the performance of the CPU can be increased by an order of magnitude. Examples of this research are the interactive simulation of material aging and the alignment of HDR image sequences.