Project 5

Abe Stephens cs6620

Required Images



Uniform Sampling 1spp
Uniform Sampling 1spp

Jittered Sampling 3x3 spp Tent filter (support 2).
Jittered Sampling 3x3 spp Tent filter (support 2).

Jittered Sampling 3x3 spp Cook filter (support 3)
Jittered Sampling 3x3 spp Cook filter (support 3)

Jittered Sampling 3x3 spp Sinc filter (support 2)
Jittered Sampling 3x3 spp Sinc filter (support 2)

Description:

The creative scene consists of a transparent glass box with one open side containing a metalic orb. The box material is a dielectric which includes a Beer's law implementation. The ground plane is a procedural checkboard texture. I distributed samples under the support of the filter and did not resuse any samples.

Performance Template

I expect the performance of my renderer could be improved by reusing samples. I rendered a slightly different scene for the Fourier images. I replaced the metalic material with a Lambertian material. I also rendered the 1024x1024 images with 16 processors on ray. It took less than an hour to render all of the images, figure out how to convert them to ascii ppm (gimp will import export), apply the transform, and add them to this web page. I didn't measure exactly how well my code scaled, however, while producing the extra images, the SGI with 16 threads beat my dual Xeon workstation with four.

Time spent: I tried to port my ray tracer to two other platforms which took a considerable amount of time- but now I can run on Irix and OSX as well as Windows. I think working out all of the bugs in the filters and constructing the creative test scene took two or three days in all. I think I should have spent more time improving the performance.

Difficulty: Medium.