High-performance visualization algorithms for large-scale scientific data.

摘要

This dissertation presents high-performance visualization algorithms for analyzing numerical data from large-scale scientific simulations. Four novel algorithms that can solve a broad range of problems in various applications of scientific visualization are included. First, we devise an efficient isosurface extraction algorithm that allows the user to analyze three-dimensional scalar fields by interactively selecting regions of constant field values. The new algorithm can save over 90% of execution time when compared with standard isosurface extraction algorithms. Second, we develop a fast direct volume rendering algorithm that can be used to visualize three-dimensional time-varying scalar fields. From our experimental studies, more than 90% of savings in rendering time and more than 85% of savings in data storage were achieved in time-varying scalar field visualization. Third, we present a global vector field visualization algorithm using the Line Integral Convolution (LIC) as the underlying method but adding dye advection to enhance local features. We present a fast searching algorithm so the LIC computation can be done efficiently. In addition, our method enables simultaneous analysis of global and local flow features. Finally, we propose a new convolution algorithm for LIC to visualize three-dimensional unsteady flow fields. Our new algorithm can produce time-accurate, highly coherent flow animations to highlight global features in unsteady flow fields. This dissertation provides solutions for many of the most challenging problems encountered in today's scientific computing/visualization environments.