Rendering competition background.
1998 was the year of the vices. Smoking, drinking, and gambling won the top three places respectively.
Patti, Kuan-loong, and Seth's outstanding images and animations took first place in the competition this year. Their showcase image of a burning cigarette included an algorithmically generated cigarette tip, volume rendered smoke, caustics from the water waves on the column, and the glossy reflection of the stone on the ashtray. They also presented two related images that featured the caustics more prominently.
The cigarette tip was modelled as a large number of concentric tobacco rings perturbed by a Gaussian noise function. The smoke is modelled by a cone perturbed by a 3D turbulence function, and the cone itself is distorted to cling to the surface of the cigarette before dissipating upwards. The caustics are rasterized onto a cylindrial "caustic map" which surrounds the object, which can be indexed when drawing the object itself. When this mapping is properly computed the caustics do not inappropriately appear on flat, horizontal surfaces pointing away from the water, such as the pedestal of the cat sculpture. The semi-diffuse (glossy) reflections are anisotropic.
More details for the interested.
Patti, Kuan-loong, and Seth's second scene was a completely algorithmically generated fractal mountain. (Although this does not fall under the heading of real world object, their primary scene does.) The first image shows the mountain from the side. The second is a bird's eye view of the entire scene, for a better view of the rivers. In the author's own (edited) words:
"The fractal mountain is modeled as a polyset created from a heightfield. A diffuse and specular texture map are algorithmically generated from the heightfield data to provide the same effect as a volume-based texture. The color scheme of the mountain was based on both height and noise to simulate rock strata. Snow is algorithmically grown on the mountain using a probability density function which takes into account the height of the point and its slope. High or flat areas are likely to have snow, while low or steep mountain faces are not. Rivers are also algorithmically grown. Rivers begin at randomly chosen nucleation points and follow the path of steepest decent to the sea. The river will pool in valleys until it climbs the walls. A corresponding specular map is also generated to add reflective properties to the snow and water but not to the mountain itself."
More details for the interested.
Michael and Chris won second place with their candlelit wine bottle scene. In the authors' own words:
"The earth is bump-mapped; in addition a modulation texture map blends the color map on the left and the emission map on the right to present the daytime and nighttime earth, complete with lights glowing in the major cities in Europe and Johannesburg, South Africa. The moon is similarly texture-mapped.
The candlesticks are texture-mapped using volumetric color texture maps. The candle flames are modeled using a combination of billboards and translucent emission mapping to blend a 2D flame image into the 3D scene, as well as a stochastically modeled, cylindrical volumetric light source. Note that the flames are visible in the reflections off the bottle.
The bottle is texture-mapped using bump, color, specularity and shininess maps, modulated by a decal map. Note the raised, reflective metallic logo, and the wine meniscus visible within the bottle. The cork is color-mapped using a different volumetric texture."
Their attention to detail was impressive: the liquid in the wine bottle even has a meniscus!
Jeremy and Matt won third place with their gambling scenes. They implemented CSG (computational solid geometry) in order to model smooth dice and cards with rounded edges. The cards were also modelled using a height field for global bending and local crumpling effects. An animation shows a card being folded, spindled, and mutilated. The backs of the cards have a subtle bump texture which comes from the card machining process. It's most noticable when the light hits the card back at grazing angles, as in the house of cards. Many of their scenes featured excellent composition, lighting, and camera positioning.
Matt, Bobby, and Brian won an honorable mention with their bat and ball on a wooden table. The ball features extensive scuffs and scratches with bumpy seams and stiches. Both the bat and ball cast glossy reflections and soft shadows on the wooden table. The brushed metal appearance of the bat is simulated with an anisotropic BRDF, more visible in the closeup image.
Mark, Matt, and Fayyaz received the Good Humor award for their scene. Their title is a cross between "peanut butter and jelly" and "Pearl Jam", the artists of the Yield CD shown. Careful observers will note that their CS348B brand of extra crunchy ray tracing peanut butter made from photons and sugar is hard to find in stores. Sadly, this peanut butter is not a significant source of vitamin A, vitamin C, or sleep. More seriously, the air bubbles in the peanut butter are created by a 3D procedural noise function. The CD is rendered with an approximation to wavelength-dependent effects. Two short animations show the effect of changing the thickness of the CD grooves and slightly changing the viewing direction.
John and Richard's scene shows blocks and marbles on a wood surface. The blocks have a subtle wood grain and "wear" textures to make them look well-used instead of brand new. Caustics formed by the translucent and semi-transparent marbles appear on the tabletop.
Sean's raytracer simulates wavelength-dependent effects, which are very noticable when observing soap bubbles. In order to get the thin-film interference effects to look right, he used spectral power distributions rather than simple RGB values for light rays. The thickness of the bubbles is perturbed by a noise function. Note that he has captured the "dark on the outside, light on the inside" look of a real soap bubble's shadow. The closeup shows the closely packed colors in a region that will average out to gray when rendered at low resolution.
More details for the interested.