What is it about?
In fog and clouds many small water droplets cause the light to scatter due to multiple refraction and reflection. Mie scattering describes this process including wave effects like interference, but is too complex to be used in realtime rendering context directly. This works takes the output of a full simulation over the visible light spectrum with respect to different droplet size distributions and develops a simple analytic function to approximate the observed scattering. The only parameter to that function is the mean particle diameter in micron for artistic freedom. Beyond simple evaluation for two given directions, we also provide an closed form algorithm to importance sample an outgoing direction for the application in path tracing.
Featured Image
Photo by Alex Machado on Unsplash
Why is it important?
Clouds and fog often appear when we render outdoor scenes for entertainment or simulation as for example to train autonomous cars. Most graphics applications adopt a simple function of one or two Henyey-Greenstein phase functions which can be a poor approximation for the actual appearance of clouds and the like. Our approximation is much more faithful while not being overly more complicated than a mixture of Henyey-Green phase functions. Our additional parameter mapping also allows to achieve a physically more plausible solution easier - without manual tweaking.
Perspectives
The actual function itself is a combination of a Henyey-Greenstein and a Drain phase function, which we found to provide the best fit over a wide range of tested combinations. True novelty does not lie in the function itself but rather in the finding that this combinations works so well. Further, our importance sampling function for the Drain phase function was not published before and can be used in other applications of the Drain function itself. While being usable with four raw input parameters we strongly advice to use the parameter fitting functions from micron to abstract parameters. While these might look spooky at first sight, you only need to compute them once when the parameter is changed. Thus they don't incur a performance overhead. Plus, it is easy to deviate from a visually good approximation to a very bad one by changing the four raw parameters uninformed.
Johannes Jendersie
NVIDIA Corp
Read the Original
This page is a summary of: An Approximate Mie Scattering Function for Fog and Cloud Rendering, August 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3587421.3595409.
You can read the full text:
Contributors
The following have contributed to this page
