Abstract
This dissertation proposes high-quality sampling methods for real-time rendering of complex effects with ray tracing. Real-time ray tracing significantly improves the visual quality in real-time rendering applications in recent years. However, the application is mostly on simple effects like reflection and shadow, which require few samples. To render complex effects like global illumination and volumetric multiple scattering, it is critical for the sampling algorithm to be intelligent to avoid generating a large number of samples to achieve acceptable quality. In this dissertation, we develop an algorithm to combine rasterization and stochastic ray tracing to render shadowed illumination from a large number of virtual point lights. We then introduce a light tree sampling algorithm to efficiently estimate illumination in general many light scenarios in fully dynamic scenes. Moreover, we propose a real-time volume rendering method based on reservoir resampling that can render heterogeneous volume with multiple scattering in dynamic, complex scenes. We further generalize the resampled importance sampling theory to support more robust spatiotemporal reuse to dramatically improve the real-time sampling quality of difficult paths in global illumination.
BibTeX
@phdthesis{DBLP:phd/us/Lin22,
author = {Daqi Lin},
title = {High-Quality Sampling for Complex Effects in Real-Time Ray Tracing},
school = {University of Utah, {USA}},
year = {2022},
biburl = {https://dblp.org/rec/phd/us/Lin22.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}