A Fast and Accurate Implementation of the Line Integral Convolution

摘要

Line Integral Convolution (LIC) is a visualization tool adequate for representing 2D flows and vector fields. The results produced by the LIC algorithm are neat and sometimes striking, clearly outperforming other techniques as streamlines, glyphs or icons. Unfortunately, speed and accuracy may be severely compromised, and there is no obvious tradeoff between these characteristics and the quality of the results. In this paper we present a technique that both enhances quality and speeds up the LIC algorithm. It is based on an adaptive sampling method that allows a geometrically uniform traverse of the streamlines, thus avoiding undersampling when the parametric velocity is high, and breaking useless computation when the streamline approaches a critical point. This technique can be easily extended to phase-shifted filter kernels to allow the representation of directional information, and pseudoanimations.