Faster ray tracing using adaptive grids and area sampling
Krzysztof S. Klimaszewski · 1995
This dissertation presents two acceleration techniques for ray tracing: the simple area sampling algorithm and the adaptive grid algorithm. The simple area sampling scheme capitalizes on the relatively low keenness of the human eye by undersampling images and interpolating the screen areas where the color variation is not precipitous. To a certain extent, the method sacrifices the accuracy of the generated images in order to increase the rendering speed up to a few hundred percent. Memory requirements are low in comparison with competing algorithms. The adaptive grid approach builds on the efficiency of the regular grid in the scenes with uniformly distributed objects. The algorithm identifies regions of such distributions and recursively applies heterogeneous grids. This isolates empty areas and equalizes voxel occupancy. The method outperforms the regular grid technique by a factor of 5 to 50 in compact scenes. In scenes with highly irregular distributions of objects the acceleration factor rises to hundreds and, if the method is combined with the area interpolator, to thousands.