A rough draft of the scene we would like to reproduce is shown in Figure 1. The centerpiece of this (only mildly contrived) still life is a clock that, at first glance, is very shiny. This seemingly simple property makes it an object of fascination to easily amused grad students. Upon close inspection, however, it becomes evident that there are a lot of interacting phenomena contributing to various aspects of the clock's shiny appearence, and we hope that modeling these in some depth will be fun (and potentially instructive).
The lighting conditions shown in these photographs are relatively flat. We acquired these pictures in late afternoon, at a time when illumination in the room was mostly diffuse. It is interesting to observe the dramatic effect on this clock of changing lighting conditions over the course of a day - an effect that just seems to be crying out for a simulated time-lapse animation. However, while the lighting conditions shown here aren't necessarily the most interesting, they will suffice to illustrate the variety of light scattering effects in the scene.
The most significant glossy/specular features in the clock
A significant feature of the scene that should not be overlooked is the rough texture of the background wall. We believe this adds considerable visual interest to the scene compared to, oh, say, a plain black background. The most challenging scene element, however, is the glass of water. In the glass itself, we see specular reflection of the environment. Next, we see complicated refractive effects as the glass and water act as a lens, distorting objects seen through the water. Finally, note how suitable lighting conditions produce bright, highly focused caustics. (see Figure 3 for a closeup of all these effects).
The final element in the scene is the newspaper. Primarily, this adds detailed texture to the scene. However, it is important to notice that the newspaper isn't completely flat ... and it would probably look quite fake if we rendered it as if it were. This shouldn't require any especially complicated modeling - our existing heightfield primitive should allow us to describe the slight non-planarity of the paper just fine.