| In this thesis, we explore several approaches to integrate compression and rendering into a single processing pipeline. The first integration scheme applies only to regular volume data sets, and performs rendering directly on compressed data sets without any decompression at all. This scheme thus reduces the data loading time and the memory footprint size, and at the same time avoids the decompression overhead completely.; The second integration applies to irregular volume data sets where volume rendering can be performed on the fly during decompression. In this scheme, the renderer can start to work immediately after the decompressor outputs the first tetrahedron, therefore the start-up latency is significantly reduced. As data sets are brought into memory in their compressed form, the data loading time is minimized. Together with a garbage collection mechanism that frees tetrahedra immediately after they have made their contributions, we further reduced the peak memory footprint size, which in turn, leads to further performance improvement. This reduction in memory foot-print size is especially helpful for those data sets whose peak memory requirements exceed the physical memory size, i.e., out-of-core rendering applications.; To address the rendering performance issue, volume simplification represents a promising approach as it provides a mechanism to trade quality for performance. However, volume simplification is mostly performed independently of volume compression. We have developed an algorithm to integrate volume simplification with volume compression, thus making it possible to perform volume simplification and rendering on the fly during volume decompression. Through this pipeline-like structure, each tetrahedron goes through the decompression stage, simplification stage, and rendering stage, if it is not simplified away. However, to be able to support “time-critical” rendering which enables interactive or even real-time volume data browsing, this integration scheme alone is not sufficient, because of the decompression overhead. To avoid decompressing a whole data set each time, we develop a multi-resolution pre-simplification mechanism to correctly determine the corresponding simplification ratio for a given frame rate, and eventually deliver the frame rate accordingly.; To demonstrate how volume compression can also be integrated into other applications as well, we have also developed an algorithm that combines volume compression with iso-surface extraction into a one-pass algorithm. (Abstract shortened by UMI.)... |
