Light transport simulation with global illumination algorithms
PR, BA, DA
Description
Various algorithms exist that are able to efficiently simulate the path of light and its behavior when hitting various different materials. With these algorithms, it is possible to render photorealistic images that are almost impossible to distinguish from physical reality. We are able to get these final images faster and faster. It is also possible to render very sophisticated physical and fluid simulations with global illumination renderer programs in a surprisingly short amount of time.
The Metropolis Light Transport algorithm is considered to be the "flagship" of the unbiased global illumination rendering algorithms, meaning that it is capable of computing exact results without any approximation involved. The Primary Sample Space Metropolis Light Transport algorithm works by creating new light paths between the camera and the light sources by introducing perturbations (modifications) to the previously evaluated light paths.
Photon mapping algorithms are biased rendering algorithms, meaning that they are expected to give results in a lower amount of render time than unbiased algorithms. This advantage comes at a cost of being an approximation algorithm, therefore most of the time the solution won't mimic physical reality "perfectly". Stochastic Progressive Photon Mapping and Progressive Photon Beams are considered to be among the finest modern photon mapping variants.
Using light groups makes it possible to adjust the settings of light sources of a scene after it has been rendered with a global illumination algorithm without re-rendering it. The usage of light groups is very easy and intuitive, therefore it gives the opportunity to an artist to completely redesign the lighting setup of a scene.
If you would like to:
- join research projects on this field,
- work with a world class, state-of-the-art physically based renderer program, LuxRender or Mitsuba,
- get your results used by a lot of people,
then this project is for you.
Tasks
The tasks of the project revolve around the following (choose one):
- implement new path building strategies to the Metropolis Light Transport algorithm to make it more efficient,
- build new ideas upon existing photon mapping-type algorithms, such as Stochastic Progressive Photon Mapping,
- implement new methods that are capable of automatically finding aesthetically pleasing lighting setups for scenes using light groups,
- and a lot more, depending on the student's personal preferences.
Requirements
An ideal candidate should:
- know basics of calculus and probability theory - i.e. should not get afraid of integrals and expected values :)
- be proficient in the C++ programming language,
- have some knowledge in ray tracing and global illumination.
