This is the final version of the scene... with many many rays being refracted through the glass pitcher on the table...

This is the final version of the scene... from another viewpoint. This lets you get a good overview of the objects the scene was constructed from.

This is what the scene looks like when rendered by the OpenGL preview mode of my raytracer. With this many patches, even OpenGL isn't 'realtime' on my poor little K6-II. :)

This is yet another view of this scene... with more complex ray interactions going on. This one was traced on a busy system, so the render time is unfairly high...

This is a close up view of the flower bouque on the corner of the table... Perdy. :)

This shows what a refractive teapot looks like. Unfortunately, this teapot model (the 'classic' one), does not have a bottom, nor a rim for the lid to sit on. Because it's not a solid model, it looks kinda funky when refracting light.

This shows the new bezier patch algorithm at work. In contrast to my previous teapot (same model, different algorithm), this one takes MANY TIMES less memory, is faster, and doesn't have cracks. I need to fix the mouse bug in the screen capture thing though. Icky.

This shows the final version of the scene, where I chose to change the table top surface. This one took a lot longer to render than the previous pictures because it is antialiased and has soft shadows. You get what you pay for. I think it turned out okay...

Once I got the objects in the scene looking nice, I had to get rid of the whole infinite plane thing. I put the museum scene on a cylindrical surface and added a 'sky' sphere with textured clouds. Of course the clouds are procedurally generated. :)

This image shows much nicer marble. :)

This image shows me trying (and failing) to create a nice marble texture for the scene... I turned off the height field in this picture to speed up rendering...

This image shows the scene from the normal perspective. Now the height field edge polygons have been fixed, and it has been placed on top of a wood block to make it taller. Note how the textures between the two match up perfectly.

This image shows a close up of the new fifth object, a height field. You can see in the image that the 'edge' polygons don't really match up correctly yet... but it looks decent (procedural wood texture map)!

This shows off the basic layout of the scene. The lighting has been preliminarily set up, and 4 of the 5 pillars have been populated. Also, the pillars have been defined as two 'block' objects, with a bunch of cylinders. The four objects are: two fog volumes (spherical and cubical), an earth globe, and an iso-surface ball.

This scene is very similar to the previous one, except now the fog in the sphere is red and much more translucent. Also, more of the sphere itself is transparent...

This scene highlights a specular sphere that bounds a fog volume... and has holes in it. Through the holes in the sphere, you can see the grey fog inside of it... through the sphere you can see the green fog volume behind it. In the solid portions of the sphere, you can see the reflections of the rest of the world in the sphere... Rendered with 3x3 adaptive supersampling.

This scene shows three objects with procedural aspects to them... the floor has a procedural bump map applied to it, the block of wood obviously has a procedural texture map on it, and the spherical object (from before) has opacity maps on it.

This scene shows a spherical object (which is really several nested spheres) that has opacity maps on it (specifically "threshold .6 noise3 scale 10 10 10" with varying thresholds for different sphere radii). It's kinda a cool effect, especially until I implement proper isosurfaces.

This image is an experiment that I did. Basically the texture map on this looks like: "texture map image.png map image.png". It is effectively taking the color that would normal be used as a texture map (which is what "map image.png" returns) and using the first two color values as indices into the texture again... causing this interesting effect.

This image illustrates an interesting bug I had to kill... this is what happened which I simply casted a double to integer, instead of using the floor function in the appropriate place. Graphics bugs are cool...

This is the final version of the image with depth of field turned on, and with a bump map applied to the table surface to make it non-perfect. I really don't like the depth of field effect, but it would have helped if I turned jittering on. :)

This version of the image shows the scene from close to the final perspective. In this version you can see the soft (non jittered) light sources casting their multilayered shadows around. This scene was engineered so that the whole table was illuminated by "soft" light, not direct light... which accounts from the major banding effects...

In this version of the scene, I just added temporal antialiasing (aka motion blur) to the cue ball... it looks like it just got hit huh? Also, again, notice the nice soft shadows on the other balls in the scene.

Here you can see the previous picture with soft shadows turned on...

This is the start of my pool scene... here you an see my cue that I constructed out of really long cones, and my table surface... the borders of the table are just brown boxes right now... but they will look nicer later.

This shows the depth of focus implementation running on a bump mapped scene...

This shows a more agressive swirly bump map on the floor. This image rendered in 31 seconds on my lowly K6-2 400.

Woo hoo, bumpmapping works! This scene has a bump mapped plane surface, a bump mapped sphere, and a "crystal ball" that is magnifying the floor... It's kinda cool how the bump map affects the reflection in the floor.

This image shows me working on bumpmapping... and it not working. You can see that the 'grooves' in the sphere don't look right. Oh well, maybe next time. :)

This is a kinda cool image that shows when I had depth of field halfway implemented... the base of the ray was being offset, but the direction wasn't being corrected. Kinda cool that you can see all the individual samples (or not? :)

This is the final version of the image... I decided to change the color of the "wine" to be red instead of yellow. This makes it clear that I know you don't drink tequilla from a wine glass. :) Also, this image is antialiased, which makes it look much nicer than the previous one...

This scene took about 65 minutes to render at full quality on a PIII-933.

This is a closeup of the stopper in the final version of the scene...

This is almost the final version of the image... here I decided to texture the stopper and the camera position is back to normal...

Here I added some nice touches to finish up the bottle... I added a stopper to the bottle and a nice looking lip. CSG is very cool.

Hrm... I want the wine glass a little higher up in the scene. I added a nice little marble block to boost it up a little bit. It's just the intersection of 6 planes... (that little square wierd thing by the neck of the bottle wasn't raytraced, it just got screen captured somehow :(

The scene is really starting to shape up with the texture maps applied! This also shows the opacity mapping that makes the square label be transparent on the non-opaque portions...

This capture shows me working on applying the small label to the neck of the bottle... notice that opacity mapping is not implemented yet... so there is a black border around the label.

The label doesn't look half bad! Nice!

In this capture, I'm trying to apply a texture map to the bottle. I scanned and touched up the texture from a real bottle... but I only wanted it to wrap halfway around a cylinder, and then trim the cylinder with a plane... obviously getting the orientation right was annoying also. :) [btw, the background is different because a diffuse surface renders faster... it just made development of the scene go faster...]

This image shows my first incarnation of the bottle, with little refinement. It was created with extensive usage of CSG operations... so it really is hollow and really is shaped pretty close to how I wanted it to.

This shows the glass after I tweaked the material properties to look more like glass... it made a BIG difference huh?

Here's my first try at getting the geometry of the glass right. It's a pretty simple object with only a few objects in it... but it looks nice!

This shows a simple CSG object... a sphere with a few smaller sphere's cut out of it. Nice huh? :)

This shows a big cubic surface (that is slicing through the smaller spheres). Sorry for the crappy resolution, it was pretty slow to raytrace (30 seconds or so) and I didn't feel like waiting 9x that long to get the full size (320x240) version...

This image shows what happens when your numerical rootfinder does not converge... in this case, I was using simple newton interation to try to converge on a large cubic... but it decided not to converge at all for part of it. I'm using sturm sequences now, which are a nice way to bound the roots and guarantee convergence. Cool bug though, huh?

This image shows texture mapping with bilinear interpolation, refraction, and antialiasing. More interestingly, it shows the interaction between the three... This scene was rendered with adaptive antialiasing, see the next image for more info...

This image shows the debugging output that I have in my raytracer for use with adaptive supersampling. Basically, every black square is sampled a single time, every green square is sampled 9 times, and the red and blue squares are sampled 3 times each. If you look at the difference between this picture and the previous one, you notice that it's doing the right thing...

This images shows texture mapping with bilinear interpolation, refraction, and antialiasing. More interestingly, it shows the interaction between the three...

This scene shows my refraction working, with the simple "pencil on a glass of water" test. Here you can also see the "modulated shadow" algorithm that I implemented: Basically if a shadow ray hits a translucent object, the light source contribution is reduced, but not completely blocked. Thus, the shadows are not completely black...