Project #5 - Ray Tracing - Math 155B

Student: Alex Kulungowski

Comments:

For my final project I implemented the following:

-- Antialiasing via jittered subpixel sampling

-- Depth-of-field via jittered eye position

-- Soft shadows (spherical light sources)

-- Illumination maps (WARNING: mixed results)

The first three techniques were more-or-less straightforward coding jobs, as
they were well documented in the text.  Illumination maps, on the other hand,
were only briefly described, and finding information on their effective
implementation and use proved difficult.  After reading about the motivation
behind the development of photon maps in Professor Jensen's book Realistic
Image Synthesis Using Photon Mapping, I'm beginning to wish I'd
concentrated on photon mapping from the start.  Because I spent so much time
wrangling with my illumination map implementation, I'm afraid my other scenes
are not very compelling, although they do demonstrate the correctness of my
implementation of antialiasing, depth-of-field, and soft shadows.

Figures:

Figure 1

  • A ray traced scene rendered using the basic algorithm:

    Figure 2

  • A ray traced scene rendered using antialiasing and depth-of-field:

    Figure 3

  • A ray traced scene rendered using antialiasing, depth-of-field, and
soft shadows.  There are three spherical lights directly behind the viewer.

    Figure 4

  • My implementation of illumination mapping follows the description in
3-D Computer Graphics.  In all of the following images, the illumination
map is the portion of the pedestal immediately behind the center of the object.
The resolution of the map below is 512x512 vertices.  I had serious problems
with noise and splotchy caustics, due in large part (I think) to the finite
nature of the mesh comprising the illumination map.  In order to guarantee a
uniform highlight (at least with the sphere), I had to increase the number of
photons (rays emanating from the light source) to an unreasonable number
(100000).  Photons are emitted from the center of the light source, even though
it is actually an extended light (i.e. a sphere).

    Figure 5

  • The same scene with a lower resolution 64x64 illumination map.

    Figure 6

  • To prove that I didn't manually produce the caustic seen above, I
rendered the same Cornell box with a variety of different shapes.

    Figure 7

  • Splotchy caustics with a lower resolution map.

    Figure 8

  • I continued to experiment with bump maps throughout the assignment,
trying (mostly in vain) to get a natural-looking wave effect.

    Figure 9

  • The torus in the following two images is a bit too bright for my taste,
but it certainly produced some interesting patterns on the illumination map!

    Figure 10