The size of volumetric datasets used in medical environments is increasing at a rapid pace. Due to excessive pre-computation and memory demanding data structures, most current approaches for volume visualization do not meet the requirements of daily clinical routine. In this diploma thesis, an approach for interactive high-quality rendering of large medical data is presented. It is based on image-order raycasting with object-order data traversal, using an optimized cache coherent memory layout. New techniques and parallelization strategies for direct volume rendering of large data on commodity hardware are presented. By using new memory efficient acceleration data structures, high-quality direct volume rendering of several hundred megabyte sized datasets at sub-second frame rates on a commodity notebook is achieved.
This is my master's thesis, carried out at the Institute of Computer Graphics and Algorithms of Vienna University of Technology. It was completed in spring 2004. My supervisors were Meister Eduard Gröller and Sören Grimm. A summary of this work has been presented at the Central European Seminar on Computer Graphics 2004 and was awarded best paper and best presentation. Parts of this work have been published in Computers & Graphics, 28 (5) and the proceedings of the IEEE/SIGGRAPH Symposium on Volume Visualization and Graphics 2004.