Am 08.03.2013 06:59, schrieb Cousin Ricky:
> clipka <ano### [at] anonymousorg> wrote:
>> Most diffuse surfaces have a brilliance quite close to 1.0. (Aside from
>> that, radiosity doesn't currently support brilliance settings other than
>> 1.0 anyway.)
>
> What about metallic surfaces?  Is there a typical value?  I find that 2.5 to 3.0
> makes a good metallic impression on surfaces with low specular reflection.

Metallic surfaces are indeed a special case. With non-metallic surfaces, 
the diffuse reflection is usually a result of scattering within the 
material itself (*), which is typically isotropic or near-isotropic, 
resulting in a lambertian or near-lambertian diffuse reflection, 
corresponding to a brilliance of 1.0.

(* As a matter of fact the classic diffuse term is actually just a rough 
approximation of subsurface scattering, presuming a near-zero 
translucency and ignoring ior.)

With metallic surfaces you do not have any scattering within the 
material - instead, any diffuse-style reflection is the result of 
repeated specular reflection at a very clustered surface. I wouldn't be 
surprised if this indeed results in quite different characteristics. My 
first guess, however, would be that those characteristics would be more 
faithfully modelled with a very rough specular highlight, rather than a 
high-brilliance diffuse term.

>> By the way, when defining new materials it is good practice to use the
>> "albedo" keyword for all diffuse, specular and phong, as this allows to
>> employ a few rules that help you get realistic results (not only in
>> HDR-lit scenes):
>>
>>     diffuse albedo D
>>     specular albedo S
>>     phong albedo P
>>     reflection { R fresnel }
>>
>> For realism, you usually want D+R<1 and S+P<R, with S+P=R/2 being a good
>> guesstimate.
>
> Wait, what?
>
> I was under the impression that S+P should be equal to R, at least for polished
> surfaces.  Is this the case with metallic reflection, or does it only apply to
> non-metallic variable reflection (such as, I presume, bright red enamel paint)?
>
> I was mere hours away from uploading years of work on metallic finishes to the
> Object Collection, and I need to know if I'm about to make any embarrassing
> blunders.

For metallic surfaces S+P=R is indeed perfectly right, so you can go ahead.

For non-metallic surfaces there is a problem because the current code 
for highlights doesn't account for the fresnel law (it does for metallic 
surfaces), so depending on the angle of incidence the resulting 
highlights will be some deal brighter than the fresnel-style 
reflections. Reducing the highlight brightness isn't a perfect solution 
for this problem (because the brightness still doesn't properly vary 
with the angle of incidence), but it's better than nothing.

Post a reply to this message