View Full Version : jensen subsurf scattering

08 August 2005, 04:33 PM
Can someone explain Henrik Jensen's dipole diffusion approximation method? I have read jensen's papers but this is the part that I can't understand very well. He seems to be placing two theoretical lights, one above and one below the surface, taking measurements from these lights to the surface and multiplies each of these distances by the extinction coefficient, the results of which go into the main equation for determining the light contribution of a sample point for a specific shading point, which is in itself a pretty complex, expensive equation to solve. What are these two lights set up in a dipole configuration doing for you? Are they set up at the point being shaded or at the irradiance sampling point?

I see other implementations out there that are simpler, like the 2003 Pixar PrMan Siggraph paper which just uses a smoothstep(), comparing the distance between the sample point and the point being shaded to the scattering distance.

What does the "diffuse dipole approximation" give you?


08 August 2005, 08:40 PM
I think the dipole diffusion approximation is just there so that you don't have to solve PDE's or do other nasty numerics in the code, it gives you speed and simplicity at the cost of correctness. It's just a made up equation designed to closely approximate the correct diffusion of light in a scattering medium. Henrik did not develop this model which is why the motivation and full description of it is not in the paper. He did give references to the original work.

I don't know if people are using this dipole approximation for CG work... but I know the images, at least the production images from PDI, in Jensen's follow-up paper (the hierarchical method) did not use the dipole approximation. I would guess that it doesn't add a lot visually for isotropic diffusion, I think you can use pretty much any ol' blurring kernel and get plausible subsurface scattering...

08 August 2005, 08:01 AM
Yeah you can use pretty much anything you like.

What I like that the dipole approximation does give you is believable colour seperation that changes non-linearly with distance. It gives a very believable effect that is hard to replicate with simpler functions.

CGTalk Moderation
08 August 2005, 08:01 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.