Comments on: SSS Node under Development http://www.blendernation.com/2007/04/21/sss-node-under-development/ Fresh Blender News, Every Day Wed, 03 Dec 2008 03:09:44 +0000 http://wordpress.org/?v=2.6.3 By: epat http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-100395 epat Tue, 08 May 2007 08:10:28 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-100395 hmm - to be honest, I can't really remember why I said it wasn't 'real' as it was a pretty long time ago when I posted that lol (I usually don't check at all once I have posted!!)! I have read Jensen's paper though and other papers on the subject and I agree that it is a very good and recognised method of approximation, used (with various slight modifications) in lots of renders with the ability to render SSS. Also, what Matt had to say is quite true. However, there are (or I know of at least one anyway) other methods for solving the radiative transport equation which I think also account for subsurface scattering in an infinite homogenous media of arbitrary shape and size which use an approximation that truncates infinate matrices and therefore is only limited by the size they are set to truncate to... assuming the user has an infinite, unlimited amount of time he could get 100% perfect, physically correct results (in fact, the results will always converge anyway so you can get results as physically correct as the floating point resolution on your computer in finite time) whereas, even if the error is set to 0, since the method brecht uses is based on the diffusion approximation you can only ever get an approiximation. Also, AFAIK the diffusion approximation only accounts for forward scattering that becomes isotropic whereas the other method accounts for all forms of the effect using exactly the same assumption that the phase function is normallised and that it depends only on the phase angle. Anyway, I might be wrong so don't take me on my word for it - I think there used to be a paper/article on it at: http://www.informaworld.com/smpp/content~content=a723730456 if you're interested although you have to register to actually read the article. I've no idea how fast this method would be as I don't think I've ever seen it implemented anywhere - the paper is pretty hard to follow and uses non-bogstandard symbols for variables as well so that could be why - but I think that the main values only have to be calulated once for a given phase function before they can be used to solve the radiative transport equation analytically so it would probably be pretty fast... -epat. :) hmm - to be honest, I can't really remember why I said it wasn't 'real' as it was a pretty long time ago when I posted that lol (I usually don't check at all once I have posted!!)! I have read Jensen's paper though and other papers on the subject and I agree that it is a very good and recognised method of approximation, used (with various slight modifications) in lots of renders with the ability to render SSS. Also, what Matt had to say is quite true. However, there are (or I know of at least one anyway) other methods for solving the radiative transport equation which I think also account for subsurface scattering in an infinite homogenous media of arbitrary shape and size which use an approximation that truncates infinate matrices and therefore is only limited by the size they are set to truncate to… assuming the user has an infinite, unlimited amount of time he could get 100% perfect, physically correct results (in fact, the results will always converge anyway so you can get results as physically correct as the floating point resolution on your computer in finite time) whereas, even if the error is set to 0, since the method brecht uses is based on the diffusion approximation you can only ever get an approiximation. Also, AFAIK the diffusion approximation only accounts for forward scattering that becomes isotropic whereas the other method accounts for all forms of the effect using exactly the same assumption that the phase function is normallised and that it depends only on the phase angle. Anyway, I might be wrong so don't take me on my word for it - I think there used to be a paper/article on it at: http://www.informaworld.com/smpp/content~content=a723730456 if you're interested although you have to register to actually read the article. I've no idea how fast this method would be as I don't think I've ever seen it implemented anywhere - the paper is pretty hard to follow and uses non-bogstandard symbols for variables as well so that could be why - but I think that the main values only have to be calulated once for a given phase function before they can be used to solve the radiative transport equation analytically so it would probably be pretty fast…
-epat. :)

]]> By: Scott http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-96431 Scott Sun, 29 Apr 2007 03:23:42 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-96431 just noticed... on graphicall, brecht has posted (on 4/25) what he says is his last build. SSS is built into the materials panel, so there is no fiddling around with nodes, there is full antialasing, back SSS, and the utah teapot .blend and a .blend with two hands are included as examples... downloaded it today, works perfectly, not too unbearably slow (at least on my computer) just noticed…

on graphicall, brecht has posted (on 4/25) what he says is his last build. SSS is built into the materials panel, so there is no fiddling around with nodes, there is full antialasing, back SSS, and the utah teapot .blend and a .blend with two hands are included as examples…

downloaded it today, works perfectly, not too unbearably slow (at least on my computer)

]]> By: Kram1032 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93254 Kram1032 Mon, 23 Apr 2007 19:15:47 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93254 argh forgot to mention that: It was meant to stop people asking, when there will be "real" SSS sry for double posting :S argh forgot to mention that:
It was meant to stop people asking, when there will be "real" SSS

sry for double posting :S

]]> By: Kram1032 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93253 Kram1032 Mon, 23 Apr 2007 19:14:51 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93253 I know ;) I didn't say, that I don't like the fake or something like that. I just wanted, to say, that biased renderers can't produce accurate SSS... SSSlow SSS is a good sign for beeing close to reality, btw :P real is really relative, at that moment at that you look at a computer-rendered picture / song / what ever. Quantum physics even say, that things from REAL LIFE don't exist, when you can't see them. Strange behaviour... I know ;)
I didn't say, that I don't like the fake or something like that.
I just wanted, to say, that biased renderers can't produce accurate SSS…

SSSlow SSS is a good sign for beeing close to reality, btw :P

real is really relative, at that moment at that you look at a computer-rendered picture / song / what ever.

Quantum physics even say, that things from REAL LIFE don't exist, when you can't see them.
Strange behaviour…

]]> By: toontje http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93175 toontje Mon, 23 Apr 2007 15:01:51 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-93175 Epat, according to Jensen's paper, this is a FAST SSS approximation where measurements showed that it is indistinguishable from real (AND OH SO SLOW) SSS calculations. So, yes, it is the real thing. Real forward scattering SSS. And if I'm not mistaken, the paper states that the more translucent (soft) the material is, the faster the calculation will be. Epat, according to Jensen's paper, this is a FAST SSS approximation where measurements showed that it is indistinguishable from real (AND OH SO SLOW) SSS calculations. So, yes, it is the real thing. Real forward scattering SSS. And if I'm not mistaken, the paper states that the more translucent (soft) the material is, the faster the calculation will be.

]]> By: ZanQdo http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92934 ZanQdo Mon, 23 Apr 2007 02:16:50 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92934 hey theres a new linux build at graphicall.org by alexv :) hey theres a new linux build at graphicall.org by alexv :)

]]> By: brecht http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92872 brecht Sun, 22 Apr 2007 22:02:52 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92872 epat, please explain what you mean by 'real' or 'proper' SSS. This method simulates a specific effect, multiple scattering under the assumption that the surface is no too thin, quite accurate and according to the BSSRDF. What effect is missing for this to be 'proper' SSS? epat, please explain what you mean by 'real' or 'proper' SSS. This method simulates a specific effect, multiple scattering under the assumption that the surface is no too thin, quite accurate and according to the BSSRDF. What effect is missing for this to be 'proper' SSS?

]]> By: Matt http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92869 Matt Sun, 22 Apr 2007 21:58:51 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92869 Kram1032, epat: Define 'real' with regards to computer graphics. Everything that's rendered from any app is an impression, a model. Even the most 'physically accurate' renderers are based on simplifications, on mathematical approximations of reality, and giving it a fancy name like 'unbiased' doesn't magically make it physically real, there are still differences in which phenomena a renderer has been coded to recreate, and what models it uses to do so. What's most important is that it gives you a perception of reality, and while some renderers use vey scientific methods of attempting that, it's not the only game in town. Brecht's method here is real in the sense that it uses Jensen's technique, but it's doing it by reconstructing the 3D surface from pixels, rather than faces. Of course this has lots of drawbacks and limitations, but from my layman's perspective at least, the technique of calculating the light interactions seems as 'real' as many others. Kram1032, epat: Define 'real' with regards to computer graphics. Everything that's rendered from any app is an impression, a model. Even the most 'physically accurate' renderers are based on simplifications, on mathematical approximations of reality, and giving it a fancy name like 'unbiased' doesn't magically make it physically real, there are still differences in which phenomena a renderer has been coded to recreate, and what models it uses to do so. What's most important is that it gives you a perception of reality, and while some renderers use vey scientific methods of attempting that, it's not the only game in town.

Brecht's method here is real in the sense that it uses Jensen's technique, but it's doing it by reconstructing the 3D surface from pixels, rather than faces. Of course this has lots of drawbacks and limitations, but from my layman's perspective at least, the technique of calculating the light interactions seems as 'real' as many others.

]]> By: epat http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92844 epat Sun, 22 Apr 2007 20:38:14 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92844 For all those wondering or mistaken - it isn't 'real' SSS - it's just using a fake method but it's very close and looks quite good! Good work! - though I'd still like to see 'proper' SSS in blender soon!! Well done brecht! -epat. :) For all those wondering or mistaken - it isn't 'real' SSS - it's just using a fake method but it's very close and looks quite good! Good work! - though I'd still like to see 'proper' SSS in blender soon!! Well done brecht!
-epat. :)

]]> By: Antiwhine http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92841 Antiwhine Sun, 22 Apr 2007 20:29:11 +0000 http://www.blendernation.com/2007/04/21/sss-node-under-development/#comment-92841 For all the people who think something doesn't work.. Read the topic linked in the article. It's all explained in there. There's no Anti-Aliasing yet, the filter will be integrated in the renderer and backwards scattering currently only works by using a cool setup with an extra camera and node. Great stuff by the way! brecht rocks! :) For all the people who think something doesn't work..

Read the topic linked in the article. It's all explained in there. There's no Anti-Aliasing yet, the filter will be integrated in the renderer and backwards scattering currently only works by using a cool setup with an extra camera and node.

Great stuff by the way! brecht rocks! :)

]]>