Hardware-Accelerated Isosurface-Based Volume Rendering of Stratigraphic Grids.

in: Proc. 29th Gocad Meeting, Nancy

Abstract

Geological models are three dimensional by nature, which calls for adapted visualization algorithms such as volume rendering. This method is generally slow because it calls for sampling the whole volume. High-performance volume visualization has been much studied on Cartesian volumes, whose simple data structure allows for drastic optimizations of visualization algorithms. However, they are unable to conform to geological horizons or faults. In contrast, curvilinear grids, also called corner-point-geometry grids, are widely used in reservoir modeling. They are suitable to model geological structures with better accuracy, but are harder to integrate in a volume rendering algorithm. Preintegrated slicing provides clear and efficient visualization for Cartesian grids. Slicing curvilinear grids is however more complex because of the irregular geometry of the mesh. We propose a new approach which considers slice boundaries as isosurfaces of a distance function D(x, y, z). Function D can be either the distance to the eye in perspective view, or the distance to a plane orthogonal to the eye direction in parallel view. Isosurface extraction is based on the Marching Cubes algorithm. Our volume rendering algorithm benefits from the recent improvements of graphics hardware. We store the model in textures for fast GPU access, which avoids data redundancy. The Marching Cubes algorithm is then performed directly on the graphic card using a geometry shader. The slices are eventually combined for color integration, which defers to a preintegrated lookup table for maximal efficiency.

Download / Links

BibTeX Reference

@INPROCEEDINGS{LaurentGM2009,
    author = { Gautier, Laurent and Viard, Thomas and Caumon, Guillaume },
     title = { Hardware-Accelerated Isosurface-Based Volume Rendering of Stratigraphic Grids. },
 booktitle = { Proc. 29th Gocad Meeting, Nancy },
      year = { 2009 },
  abstract = { Geological models are three dimensional by nature, which calls for adapted visualization algorithms such as volume rendering. This method is generally slow because it calls for sampling the whole volume. High-performance volume visualization has been much studied on Cartesian volumes, whose simple data structure allows for drastic optimizations of visualization algorithms. However, they are unable to conform to geological horizons or faults. In contrast, curvilinear grids, also called corner-point-geometry grids, are widely used in reservoir modeling. They are suitable to model geological structures with better accuracy, but are harder to integrate in a volume rendering algorithm. Preintegrated slicing provides clear and efficient visualization for Cartesian grids. Slicing curvilinear grids is however more complex because of the irregular geometry of the mesh. We propose a new approach which considers slice boundaries as isosurfaces of a distance function D(x, y, z). Function D can be either the distance to the eye in perspective view, or the distance to a plane orthogonal to the eye direction in parallel view. Isosurface extraction is based on the Marching Cubes algorithm. Our volume rendering algorithm benefits from the recent improvements of graphics hardware. We store the model in textures for fast GPU access, which avoids data redundancy. The Marching Cubes algorithm is then performed directly on the graphic card using a geometry shader. The slices are eventually combined for color integration, which defers to a preintegrated lookup table for maximal efficiency. }
}