Rendering Speed Improvement for Particle Based Fluid Simulation on GVDB Voxels using Whitted Raytracing

Fluid simulation is the representation of fluids and its displacement as in the real world on digital media such as computers. This research specifically discuss problem regarding visual rendering for fluid simulation. A common approach to represent fluid simulation is through groups of point representing volumetric fluid particles. This data representation approach is called as point cloud data. GVDB Voxels is a new framework that may aid storing, calculating, and rendering of fluid simulation data stored with the framework. In this research, it is specifically discussed techniques to render fluids stored in GVDB Voxels. This research explains how point cloud data is stored in GVDB Voxels, converted into level set data, rendered to images using raytracing, and lastly modified to perform trade-off between render speed and render quality. Raytracing was performed using NVIDIA OptiX Engine integration with GVDB Voxels. Using OptiX Engine, we implement a Whitted Ray tracer as alternative to the provided Monte Carlo Path Tracer. Moreover, we propose Quincunx antialiasing method to deal with appearing aliasing problem. Rendering with the proposed solution is able to reach realtime performance with FPS count 30.47 and higher SNR (signal-to-noise) value than the provided Monte Carlo Path Tracing method in particular cases. Future work related with this topic should aim to modify secondary ray tracer by incorporating depth bounding which may generate for FPS gains.