Mike Stark
Programming Assignment #1
Monte Carlo Radiosity Solution, Simple Mesh
Description of the Results
The images below were created using rectangular meshes, with piecewise
linear basis functions, using the dirac delta function at each of the
nodes for the duals. The mesh was constructed by refinement until the
largest dimension of each mesh element was no larger than a limiting
size, in world units. (The total height of the enclosed box is 548.8
units; the stopping points were 100 and 10 units as labeled in the
table below.)
I apparently introduced a bug in my Monte Carlo radiance function
which causes several of the objects to appear partially transparent.
Obvisouly, I haven't gotten to fixing it yet! Actually, maybe I'll
leave it. it looks kinda cool.
|
4 Samples per Mesh Node |
100 Samples per Mesh Node |
| 100 unit cells |
| 
|
| 10 unit cells |
| 
|
"One of these things just doesn't belong here!"
Originally I had an error in my mesh construction which caused certain
mesh points which should be shared to be computed independently. The
image at the lower left was constructed using this version of the
code--notice that there are discontinuities at the mesh edges.
The repaired version is tracing as I'm writing this.
Conclusion
It's apparent from the images that increasing the number of samples
per node is more effective tha refining the mesh. This certainly
stands to reason, because a good approximation to a bad representation
doesn't really get you much.