Ph.D. Dissertation Proposal
Real-time Spectral Rendering of Atmospheric Optical Phenomena
Ari Blenkhorn
2:00pm Wednesday, 10 June 2015, ITE 352
Glories, rainbows, and coronas are colorful atmospheric effects which occur when sunlight interacts with cloud droplets. Adding these effects to digital cloud environments will provide increased realism and a greater sense of immersion. Furthermore, these phenomena are the subject of active scientific research. In both communities, high-resolution real-time rendering is desirable.
The color distribution of these phenomena is typically calculated using the Mie scattering theory, Debye series, or Airy theory. The calculations give the intensity of a single wavelength of light at a single scattering angle. They must be repeated for all desired wavelengths at all desired pixels of the final image.
I propose accelerating the calculations by using general-purpose GPU computing to transform a single-threaded, CPU-based Mie scattering application into a collection of highly-parallel GPU calculations. I also propose to reduce the number of wavelengths required by using importance sampling, a monte-carlo selection method which concentrates the computing resources on the wavelengths belonging to the most important regions of the visible spectrum.
Planned work includes development of both numerical and perceptually-based image quality metrics, of interest to optical physicists and interactive application developers, respectively. These metrics will guide development of the GPU kernel parallel structure and the selection of a suitable estimator for importance sampling.
Committee: Drs. Marc Olano, Penny Rheingans, Curtis Menyuk, Matthias Gobbert (Mathematics), Raymond Lee (USNA)