Interactive and perceptually enhanced visualization of large, complex line-based datasets

Increasing computing power, including that available from Linux clusters and supercomputers permits increasing accuracy in modeling and physics-based simulation of natural phenomena, which, in turn, results in increasingly larger and more detailed data. One class of this detailed data is consists of sets of different line types which should be visualized simultaneously to understand their relationships to one another. Examples include electric and magnetic field lines, velocity and vorticity field lines, or the trajectory paths of different particle types. The extreme size and high density of these datasets can overwhelm current visualization techniques by requiring too much memory, or too much compute time. Very dense lines can also result in unintelligible images where the depth relationships between the lines can not be conveyed with current methods. To circumvent the pitfalls of current visualization technique with respect to difficult line-based data, the visualization techniques presented in this dissertation take into account perceptual issues. By providing the cues that the human visual system uses to resolve depth relationships, making use of the natural image processing capabilities of that system, and simulating real-world illumination, the visualization methods avoid ambiguity and present compelling images which are easily interpreted, and which reduce the likelihood of illusions. Three visualization methods are presented, each with their own area of applicability. The first method efficiently combines two existing line rendering techniques, and is suitable for older machines with slower CPUs and graphics hardware with minimal texturing capabilities. The second method provides self-orienting geometry for texturing. It is appropriate for large datasets where the lines to be visualized project to be at least several pixels wide, and presents lines with the appearance of three dimensional tubes, but much more efficiently than using actual polygonal tubes, and with a greater degree of flexibility for taking full advantage of the sophisticated texturing capabilities of modern graphics hardware. The final visualization method is based on anisotropic volume rendering, and is appropriate for gigantic, very dense line datasets, where line thicknesses can range from one pixel to extremely sub-pixel. Together, these three methods cover a very broad range of dataset sizes, and are demonstrated for visualizing simulations of wake vortices around the tail of an airplane, electric and magnetic fields within a linear accelerator structures, and charged particle trajectories generated and influenced by those fields.