GPU ray casting of virtual globes
Analytical Graphics, Inc.
ACM SIGGRAPH 2010 Posters, SIGGRAPH ’10
@conference{cozzi2010gpu,
title={GPU ray casting of virtual globes},
author={Cozzi, P. and Stoner, F.},
booktitle={ACM SIGGRAPH 2010 Posters},
pages={1},
year={2010},
organization={ACM}
}
Accurately rendering an ellipsoid is a fundamental problem for virtual globes in GIS and aerospace applications where the Earth’s standard reference surface is non-spherical. The traditional approach of tessellating an ellipsoid into triangles and rendering via rasterization has several drawbacks [Miller and Gaskins 2009]. Geodetic grid tessellations oversample at the poles (2a), which leads to shading artifacts and ineffective culling. Tessellations based on subdividing an inscribed platonic solid lead to problematic triangles crossing the International Date Line and poles (2b). We present a new approach to globe rendering based on GPU ray casting. Instead of tessellating the ellipsoid, we treat it naturally as an implicit surface. Simple proxy geometry bounding the ellipsoid from the viewer’s perspective is rendered in order to invoke a fragment shader that casts a ray to find the ellipsoid’s visible surface and shade accordingly. Our approach has the traditional advantages of ray casting implicit surfaces: infinite level of detail, trivial memory requirements, and simplicity. Furthermore, our approach reduces ray misses, runs at real-time frame rates on commodity GPUs, and easily integrates into existing rasterization-based engines.
February 27, 2011 by hgpu