Per-Fragment Image-based Flow Visualization using Interactive Surface Extraction from Unstructured Grids

Abstract

We present a method for image-based visualization of flow fields on unstructured grids. For interactive explorationof flow data from CFD simulations, we combine GPU-accelerated surface extraction methods and line integralconvolution (LIC) in a multi-pass algorithm. In contrast to other highly efficient methods, computation of streamlines is performed directly on the unstructured grid, where in previous methods the flow field had to be confinedto the extracted surface. The introduced algorithm is based on three passes: Surface extraction, rendering, andcomputation of stream lines to perform line integral convolution. Point location, a major performance factorin stream line computation on unstructured grids, can be greatly accelerated by reusing data from the surfaceextraction pass. This allows us to achieve interactive frame rates on current generation graphics hardware for thepost-processing of unstructured CFD datasets.