Ray tracing techniques for computer games and isosurface visualization
Ray tracing is a powerful image synthesis technique, that has been used for high-quality offline rendering since decades. In recent years, this technique has become more important for realtime applications, but still plays only a minor role in many areas. Some of the reasons are that ray tracing is compute intensive and has to rely on preprocessed data structures to achieve fast performance. This dissertation investigates methods to broaden the applicability of ray tracing and is divided into two parts. The first part explores the opportunities offered by ray tracing based game technology in the context of current and expected future performance levels. In this regard, novel methods are developed to efficiently support certain kinds of dynamic scenes, while avoiding the burden to fully recompute the required data structures. Furthermore, todays ray tracing performance levels are below what is needed for 3D games. Therefore, the multi-core CPU of the Playstation 3 is investigated, and an optimized ray tracing architecture presented to take steps towards the required performance. In part two, the focus shifts to isosurface raytracing. Isosurfaces are particularly important to understand the distribution of certain values in volumetric data. Since the structure of volumetric data sets is diverse, op- timized algorithms and data structures are developed for rectilinear as well as unstructured data sets which allow for realtime rendering of isosurfaces including advanced shading and visualization effects. This also includes tech- niques for out-of-core and time-varying data sets.