Compact iso-surface representation and compression for fluid phenomena

Abstract

We propose a novel method of compressing a fluid effect for realtime playback by using a compact mathematical representation of the spatio-temporal fluid surface. To create the surface representation we use as input a set of fluid meshes from standard techniques along with the simulation's surface velocity to construct a spatially adaptive and temporally coherent Lagrangian least-squares representation of the surface. We then compress the Lagrangian point data using a technique called Fourier extensions for further compression gains. The resulting surface is easily decompressed and amenable to being evaluated in parallel. We demonstrate real-time and interactive decompression and meshing of surfaces using a dual-contouring method that efficiently uses the decompressed particle data and least-squares representation to create a view dependent triangulation.