Physically based rendering of volumetric data plays a vital role in contemporary science, medicine, and various industries. However, existing solutions often suffer from non-interactivity and are platform-specific, constraining their practicality. In this study, we introduce a novel approach combining the principles of flux-limited diffusion approximation and multigrid method. This enables us to create an interactive global illumination simulation, regardless of the camera's position. We implemented this technique within a web application, leveraging the WebGL 2.0 interface, and conducted a performance and quality assessment. Our findings indicate that the multi-grid method significantly enhances convergence speed, and its increased memory consumption can be managed through lower resolution caching. While our approach may not match the quality achieved by path tracking methods, the diffusion approximation allows for real-time interaction with the camera, and, in less complex scenarios, facilitates real-time interaction with lighting and transfer functions.
|
