Interactive GPU-based Volume Rendering

Abstract: Modern GPUs possess enormous computational capabilities, power that has just become available for general purpose programming via the development of highly programmable shaders. This thesis investigates ways to use this power for the purpose of medical visualization. The project was carried out in collaboration with Sectra Imtec, a Linköpingbased medical imaging company. The problem to be solved was how best to render high quality images at interactive frame rates for the purpose of previewing a medical data set. Two different GPU-accelerated volume rendering approaches were investigated and implemented. One approach based on textured slices of proxy geometry, and one based on ray casting. It is shown that the ray casting implementation presented far superior image quality. Because of this, most of the work during the project concerned improving the method enough to make the visualization interactive. In particular, an empty space skipping method via a complex bounding geometry was implemented and is described in detail. This report also presents important background facts regarding volume rendering and modern graphics cards, so that it is accessible for any reader with basic computer graphics knowledge.