Here are some rather nice pictures I made with the raytracer I wrote for 15-462
in Fall 2004.
Recursive reflection test. The first of my half-decent
tests. One of the more interesting things here is that the small sphere
in the center (which is actually very much closer than the others) has only
one specular highlight, while the others have two. This is because the light
over your left shoulder is yellow, and the center sphere is only specular in
blue, whereas the light over your right shoulder is white.
Transmission test. An interesting thing to notice
here is the very small yellow shadow at the bottom left of the large gray
sphere. It is the colored shadow through the yellow triangle from a rather
bright white light a long way off the top-right of the screen. The triangle
looks weird because there is another light in the middle of the scene.
Blinn object test. My first test of Blinn objects.
Because they are numerically approximated rather than solved in closed form,
Blinn objects take very much longer to render than triangles and spheres. As an
example, the previous tests take on the order of seconds to render with
recursion depth 6 and 4x4 antialiasing; this takes about 15 minutes. The Blinn
object looks very unreal because it is highly reflective, highly transmissive,
and somewhat diffuse, which means it gives off a little over twice as much
light as falls on it.
My final demo scene. This takes on the order of 45
minutes to render on unix40 during the summer when no one is using it. The
Earth in the background is pretty cheap; the texturemapping on a sphere is
just inverse Mercator projection with the seam forced to the -z direction
(in back), so I just grabbed a standard satellite map from the web and stuck
it on there. The sphere inside the red object looks a little lame because it
is a near-perfect mirror (just like the one a little down-right of the red
object), but all it can see is the inside of the red object. Also, the weird
non-shaded circle in the shadows on the right is caused because one of the
light sources is actually aimed directly through the neck of the green object,
so there's a region of double-shadow in the middle of a region of
quadruple-shadow. I didn't feel like implementing volumetric shadows, although
if I did real refraction I'd probably do that too (which would be pretty cool).
