CS 667 Project 2: Photon mapping

Your mission is to add photon mapping to a ray tracer. You can use the same code base that you used for the radiosity assignment, or any other ray tracer that doesn't already do photon mapping or nontrivial scattering media is fair game. The due date is April 1.

I've decided to have you work in a single team again, because some of the more challenging features are too much fun to pass up. There should be plenty here to keep you all busy!

The requirements are:

• All transport paths accounted for via surfaces or volume scattering.
• Caustics rendered sharply using a photon map.
• Indirect illumination rendered using a gather from a photon map, with irradiance caching for performance.
• Scattering media rendered using ray marching for single scattering and a volume photon map for multiple scattering
• Surface materials: Lambertian; dielectric with absorption.
• Volume materials: absorption and isotropic scattering but no emission or nontrivial phase functions.
• Surfaces: spheres; triangles; triangle meshes; volumetric implicit surfaces (that is, the zero level set of a scalar function defined on a 3D grid).
• Volumes: uniform media; volumetric density (that is, scattering and absorptioin coefficients proportional to a scalar function defined on a 3D grid).

Here are the still images you should render to demonstrate your system:

• A Cornell Box with a metal sphere and a glass sphere, illuminated by its own light source.
• A Happy Buddha (data available here) made out of colored glass, standing on a surface and illuminated by a small key light above and two large (but dim) rectangular lights on either side for fill and for surface highlights.
• A flashlight beam in an aquarium full of weakly scattering medium (cf. Bohren Fig. 14-2).
• The same beam in an aquarium full of more strongly scattering medium (cf. Bohren Fig. 14-3).
• Billowing black smoke from this data (high absorption, low scattering). You should be able to see a shadow cast by the smoke.
• Billowing white steam from the same data (high scattering, low absorption). You should be able to see a shadow cast by the steam.
• Pouring water from this data, pouring into this glass (no, that one is wrong, use the two files here: outside surface; inside surface. Use lighting similar to the Buddha scene.
• The water rendered as apple jiuce (moderate brownish absorption).
• The water rendered as milk (maybe take coefficients from this paper).

Implementation hints

You will likely find it helpful to have every volume attached to a "clipping surface" so that the volume only exists inside the volume defined by that surface. You will need to clip the water surface to the inside of the glass. Also note that you will need some kind of mechanism by which a ray, when it exits the glass, figures out whether it is going into water or into air.

The data formats are unfortunately many and varied. The smoke data is in a self-explanatory (I hope) text format, and the Buddha triangle mesh is in the CS417 XML format. The water glass is a triangle mesh in the ASCII Ply format, which shouldn't be too hard to figure out. If you're interested, you can also see the Bézier spline from which it was made (in Adobe Illustrator format, which is actually nicely documented)

The water data is in a simple binary format that is read by the code in levelset.cc. This code (hastily written though it is) can also serve as an example of how to interpolate volume data and how to intersect rays with the level set. Feel free to cut and paste as much as you like into your program. Note that the water data has the special property that the numbers in the volume represent the distance to the closest point on the surface, which enables considerable acceleration by taking big steps when far away.

To make the glass fit the water in the past, I have used the transformation:

    translate 0 46 0
    scale 0.005 0.005 0.005
    translate 0 -1 0