This is an archive of a past semester of this course. Go to the current semester.
Final Project Ideas - Rendering

Rendering Volumetric Scattering (Difficulty Level: 2.5)

In project 3, you implemented surface rendering where it is assumed that light only scatters when it hits a surface. In this case, the rendering equation is an integral over all surfaces in the scene. However, this surface rendering technique is not able to render some cool volumetric scattering effects like the fog shown above.

To model volumetric scattering, you need to compute an integral over all volumes and all surfaces in the scene. You can do this by modifying your path tracer. The main difference is that a ray may get scattered before it hits any surface in volume. You may find the following resourses useful for the project.

This chapter of Wojciech Jarosz's thesis introduces some basic concepts about volumetric scattering. This paper by Lafortune and Willems has implementation details. You can skip section 4 of the paper for a simple path tracing implementation.

Progressive Photon Mapping (Difficulty Level: 2.5)

You may have noticed that the path tracer in project 3 is very inefficient when rendering caustics. To improve caustics rendering, you can implement a technique called photon mapping. Photon mapping is more efficient on caustics because it allows path samples to be shared across pixels. The core part of photon mapping is a stucture to lookup photons inside a sphere in the scene. Options for the structure include KD-tree and hash grid. Chapter 15.6 in PBRT is a good reference on photon mapping.

However, photon mapping is a biased method! Intuitively, it will always give you a blurred shading result, since the shading point is using the photons from its neighborhood. You need to consult this paper on progressive photon mapping to improve it.

Subsurface Scattering (Difficulty Level: 2.0)

Here is the seminal 2001 paper with a dipole approximation. Newer methods have greatly improved on this, but it's a good place to start. This has lots of complicated math, but once you understand what it is trying to do, it will be much easier to implement.

Hair Rendering (Difficulty Level: 2.5)

Check out Marschner et al., the paper cited in class. You may also find this paper useful. You will face a tough problem if you want to ray trace hair curves. So we recommend you use Mitsuba Renderer, where ray-hair intersection is already implemented.

You're also welcome to implement Lingqi's state of the art fur reflectance model.

Rendering Glints (Difficulty Level: 2.5)

You can add textures to your path tracer too! After that, you can extend the microfacet model to support glints! Checkout this paper and this paper for guidance on an implementation of glints using high resolution normal maps. You're also welcome to implement the stochastic glints model without normal maps.

Here is a guide to doing this with ray differentials (a slightly offset ray traced next to your main ray that lets you calculate mipmap levels). You will need it to compute a pixel's coverage on the texture.

Cloth Rendering (Difficulty Level: 1.5)

We know that the cloth can be rendered in three ways: as actual fibers, as participating media and as a surface. Here, we recommend you a simple surface model to render more interesting cloth.

Scattering Effects in a Participating Media (Difficulty Level: 2.0)

This paper offers a physically based approach to rendering fast and efficient atmospheric scattering effects in real time for isotrophic point lights.