My dissertation investigates the effectiveness of Dynamic Diffuse Global Illumination (DDGI) and explores how it can be enhanced for real-time applications. This research is implemented directly within Spark, utilizing my custom Vulkan architecture as a technical foundation. A primary focus of this investigation is the integration of DDGI directly into ReSTIR DI to analyze potential improvements in sampling efficiency and lighting stability.

I am also exploring how Screen Space Global Illumination (SSGI) can be coupled with DDGI to resolve high-frequency details that probe-based solutions often miss. To handle the noise inherent in these stochastic methods, I am integrating NVIDIA DLSS Ray Reconstruction as the primary denoising solution. To rigorously evaluate these techniques, I am implementing a reference Path Traced Global Illumination solution. This allows for a direct visual and performance comparison between the hybrid DDGI implementation and a ground truth renderer, providing clear metrics on the fidelity and efficiency of these combined techniques.

Dissertation Project