Graduation Thesis
This project is my graduation thesis for the Bachelor's degree in Digital Arts and Entertainment (Game Development major) at Howest Hogeschool West-Vlaanderen at Kortrijk, Belgium. The research compares two common approaches for volumetric lighting in games: single scattering (SS) and multiple scattering (MS). The focus of the comparison is on how the different volumetric lighting methods affect the visual quality and realism of the scene, combining performance benchmarking and a small user study.
Abstract
Volumetric lighting is often used in games to improve the atmosphere of a scene, but it is also expensive to render. Because of this, developers usually must choose between different approximations to balance the realism needed with the performance impact. This thesis compares two common approaches for volumetric lighting: single scattering (SS) and multiple scattering (MS). Both techniques are implemented in the same custom deferred renderer using Vulkan. The focus of the comparison is on how the different volumetric lighting methods affect the visual quality of the scene. The comparison combines performance benchmarking and a small user study. Performance is evaluated by measuring frame time and FPS under low and high fog densities. Visual quality is evaluated through an online questionnaire (N=34), where participants rate realism and visual appeal of paired renders (SS vs MS) and provide optional comments. Results show that the performance difference between SS and MS is negligible in the current implementation (+0.3% FPS for low fog density and −0.8% FPS for high fog density). In the user study, SS received more consistent realism and appeal ratings for high-density fog, while MS produced more polarized responses. Overall, the findings suggest that the tested MS approximation does not consistently improve perceived visual quality, even though its runtime cost remains similar to SS.