My dissertation investigates the unification of modern real-time rendering techniques to achieve near path-traced fidelity. A major technical component of this research is its implementation within a custom Vulkan renderer that I developed entirely in my free time. This personal engine serves as the foundation to explore the seamless integration of Dynamic Diffuse Global Illumination (DDGI) directly into ReSTIR DI. By treating DDGI probes as virtual light sources and utilizing unified resampling, I evaluate how these combined stochastic methods optimize direct shadowing and indirect irradiance gathering.

To push visual quality further, my work analyzes secondary vertex sampling to improve occluded shading alongside Screen Space Global Illumination (SSGI) to resolve high-frequency details that probe-based solutions often miss. Because unifying these complex algorithms introduces significant stochastic noise, I am integrating NVIDIA DLSS Ray Reconstruction as the primary neural denoising solution.

I have also developed a reference Path Traced Global Illumination solution. This provides a ground truth for comparing the performance, sampling efficiency, and visual stability of my hybrid ReSTIR and DDGI implementation against offline rendering standards.

Dissertation Project