Radiance Caching for Participating Media
(Image Comparison)
Below you can find image comparisons for the four different scenes used in the paper "Radiance Caching for Participating Media". We compare our new volumetric radiance caching method (VRC) to volumetric photon mapping and path tracing. The images in each set are equal render time.
Tip: Move the mouse over the images to compare the algorithms on specific parts of the scene.
Warning: This page include many full-resolution images, so the page may take some time to fully load.
◄ back to the publication web page
Jump to scene:
Cars
Two cars in a dense fog on a road illuminated by 60 lights. Our radiance caching method renders this scene in 20.35 minutes using only 175K cache points and 27M cache queries. The multiple scattering computation traces 4K random-walk paths per cache point. The images were rendered with a horizontal resolution of 1K with up to 16 samples per pixel and the error tolerances were set at 0.25 and 0.75 for the single and multiple scattering caches respectively. We compare to both path tracing and photon mapping. Due to the large scene extent and many light sources, artifacts remain in the homogeneous medium even after tracing 8M photons for photon mapping.Single Scattering
Single+Surface+Multiple Scattering
Cornell Box
Below is a series of smoke-filled Cornell boxes showing our method’s ability to handle isotropic and anisotropic scattering in homogeneous or heterogeneous media. All images were rendered at 1K×1K with up to 16 samples per pixel and include single scattering from lights, single scattering from surfaces, and multiple scattering. We report render times with each image and show the number of cache points in the single, surface, and multiple scattering caches in parenthesis.Isotropic Scattering
Anisotropic Scattering
Smoke
A still frame from an animation of heterogeneous smoke renders in 5.8 minutes using 9K cache points and 8.2M cache queries. Our method is able to produces flicker-free animation even when recomputing a new cache for each frame. Error tolerances of 0.05, 1.0, and 1.0 were used for the single, surface, and multiple scattering caches with 2K multiple scattering paths per cache point.Single Scattering
Single+Surface Scattering
Single+Surface+Multiple Scattering
Sponza
The Sponza atrium with beams of light and multiple scattering rendered at a horizontal resolution of 1K with up to 16 samples per pixel. Our method renders this scene in 19 minutes using about 46K cache points and 79M cache queries (error tolerances set at 0.5 and 0.01 for the surface and multiple scattering caches respectively). For multiple scattering, our method traces 512 random-walk paths per cache point.
