Particle-Based Anisotropic Sampling for Two-Dimensional Tensor Field Visualization

Abstract

We present a particle-based approach to generate unstructured distributions of elliptical samples. Size and shape of the samples are determined by a local metric that is derived from a two-dimensional tensor field. In contrast to previous methods, we propose the use of an anisotropic Delaunay triangulation of particle positions. It guarantees exact neighbor computations and provides a good means for an explicit and automatic control of prominent holes and overlaps, which otherwise would result in unpleasant visualizations. We use the final distribution to compute a generalized Voronoi diagram, which represents a novel and flexible visualization technique for two-dimensional tensor fields. Via texturing of Voronoi regions, many possibilities arise to design the final image.