| In this dissertation, we show the results of our research on the field of volumetric rendering of unstructured grid data sets. We present and explain in detail our most important results and contributions.; Volumetric rendering is a highly computational intense process. Images generated by this process show informations about the interior of the data, not only about the surface, by considering the data composed by semi-transparent materials.; A technique to be used as a tool for real-time analysis, is required to achieve a minimum rate of 10 frames per second (ideally 30 fps). Even the fastest algorithm for volumetric rendering takes about 3.5 seconds to generate an image of a data set composed by half million cells in a computer with a fast processor, nowadays.; In the first part of our research, we optimized the fastest algorithm (at the time), developed by Bunyk et al. [6], and put together some approximation techniques to speed up the image generation. In a time-critical fashion, approximate images are delivered until the system is given more time, when it generates more and more accurate images up to the exact one.; In the second part, we present a novel, simple, fast, memory efficient, exact and robust volume rendering algorithm based on the sweep paradigm, called ZSweep. This algorithm and the further work we developed on it, became the most important contribution of our research. Extensions developed on this algorithm are its adaptation to run on shared memory architectures, results got on SGI machines, and time and memory requirements of an out-of-core implementation. Also another simple out of core volume rendering algorithm is proposed. |
