Multigridding Hierarchical Radiosity
Programming Assignment #5
Simon Premoze
1. Cornell Box -- Constant Basis
After Iteration 1: 3210 elements, 10603 links, 760 Kb
After Iteration 2: 5219 elements, 38319 links, 1718 Kb
After Iteration 5: 6639 elements, 55545 links, 2704 Kb
After Iteration 5
Back view
We take advantage of multigridding method to reduce number of interactions.
First, set of links are created using a large error bound. After solving
the system, we know where most of the error is and we further refine the
links with a lot of error. After each iteration, the tolerance is halved(i.e.
tolerance = tolerance / 2.0 ). I also experimented with some other
decrements (square root, log), but it did not change much the overall performance. The system converged after 4 iterations.
Simon Premoze
premoze@cs.utah.edu