Isosurface ambient occlusion and soft shadows with filterable occlusion maps

Abstract

Volumetric data sets are often examined by displaying isosurfaces, surfaces where the data or function takes on a given value. We propose a new method for rendering isosurfaces at interactive rates while supporting dynamic ambient occlusion and/or soft shadows and requiring minimal pre-computation time. By approximating the occlusion in a region as the percentage of occluding voxels in that region, we reduce the ambient occlusion problem to the same problem faced in soft shadow mapping algorithms. In order to quickly extract the number of occluding voxels in an image region, we propose representing distributions using filterable representations such as variance shadow maps or convolution shadow maps. By choosing different sampling patterns from these maps we can dynamically approximate ambient occlusion and/or soft shadows.