Flow visualization via partial differential equations

Abstract

The visualization of stationary and time-dependent flow is an important and challenging topic in scientific visualization. Its aim is to represent transport phenomena governed by vector fields in an intuitively understandable way. In this paper, we review the use of methods based on partial differential equations (PDEs) to post-process flow datasets for the purpose of visualization. This connects flow visualization with image processing and mathematical multi-scale models. We introduce the concepts of flow operators and scale-space and explain their use in modeling post processing methods for flow data. Based on this framework, we present several classes of PDE-based visualization methods: anisotropic linear diffusion for stationary flow; transport and diffusion for non-stationary flow; continuous clustering based on phase-separation; and an algebraic clustering of a matrix-encoded flow operator. We illustrate the presented classes of methods with results obtained from concrete flow applications, using datasets in 2D, flows on curved surfaces, and volumetric 3D fields.