New volume reconstruction and parallel indirect volume rendering methods (Parallel processing, Isosurface extractions)

This dissertation investigates issues in reconstructing volume data and in visualizing volume data. The primary area of investigation in volume reconstruction is the recovery of a volume of data from a limited number of 2D views of that data. In particular, a new hybrid technique for reconstructing the plasma distribution around the Earth from satellite images is presented. The primary area of investigation in volume visualization is efficient indirect volume rendering. One major focus in this area is study and development of efficient MIMD parallel out-of-core approaches for large dataset isosurface extraction in multiprocessing environments. The approaches consider reduction of memory requirements and optimization of disk I/O, as well as load balancing across processors for the isosurfacing. Another focus in this area is study and development of efficient multithreaded in-core and out-of-core isosurfacing approaches on new desktop (single- and dual-CPU) computers with hyperthreading. The approaches consider reduction of memory requirements and total performance of isosurfacing. Experimental results and analysis are presented to validate the effectiveness of the approaches developed in the dissertation.