When I was in the hospital, I had nothing to do. I used my notebook to watch my papers every day. I turned to a Siggraph 2009 poster named "re engineering ection model of metallic paints for Re engineering ectance acquisition". written by Masashi Baba and Naoki Asada at Hiroshima City University. Tests show that their model is a little more complex than mi_metallic_paint, but the flake layer is different from mi_car_paint. Here, the flake layer uses Gaussian distribution, that is, the sample is used to disturb the second highlight. I directly sample the CDF $ \ frac {1} {2} [1 + ERF (\ frac {X-\ Mu} {\ sgima \ SQRT {2} of Gaussian distribution in the shader} })] $ to obtain samples.
When I was in hospers, I spent my on bed time by reading papers, and I found a Siggraph 2009 poster named "re warning ection model of metallic paints for RE warning ectance acquisition" which is written by Masashi Baba and Naoki Asada from Hiroshima City University. their model is fast, simple, similar to the mi_metallic_paint, but the flake model is different than mi_car_paint.
Samples = 128, Sigma = 0.149, Mu = 0.001
Samples = 128, Sigma = 0.249, Mu = 0.001
Samples = 64, Sigma = 0.349, Mu = 0.001
Download from ortholab