First Images

The initial development of the raytracer revolved around intersecting rays with geometry in a scene. The first render lacked shading and was simply a ray-intersection test on a sphere and a plane. Soon afterwards I adopted the phong-shading model and made this image.

Phong shading, no shadows.

These type of renders served as good sanity tests thanks to their simplicity and render speed. Most early changes ended up coming back to this image to make sure I didn't break anything.









Then I added shadows and a new shading model: the Cook-Torrance model. The Cook-Torrance model uses the wavelength of light to approximate specular lighting. However my raytracer doesn't use wavelengths, it uses RGB colors to simulate light. I solved this by substituting a separate equation to approximate the term.

Cook-Torrance shading relies on a different lighting model for one of its terms. In their original paper, Cook and Torrance suggest Beckman shading. However I noticed for very smooth surfaces that Beckman shading breaks down and produces sloppy results. So for smooth surfaces I substituted in Gaussian shading to approximate the term.

Shadows and Cook-Torrence shading.

Here I used a smooth surface on the sphere to show the Gaussian shading.


Notice the 2 shadows and the oblong specular highlight caused by 2 light sources.

A rough sphere to demonstrate Beckmann influence on
Cook-Torrance shading.

The two images aren't too different. The differences between Gaussian and Beckmann influences on the Cook-Torrance model are subtle.













That covers the simple stuff for the Raytracer. Next is reflections and refractions and bounding volume hierarchy.