Jim Henderson wrote:
> Ah, so you've decided Darren and I were right after all, then? ;-)

yes, now I'm a satanist. :P </sarcasm>

Post a reply to this message

Orchid XP v7 wrote:
> As best I can tell, the algorithm described just sounds like POV-Ray's 
> radiosity with an infinitely low error_bound. (I.e., always resample.) 
> But applied to *all* terms, not just diffuse illumination...
> 

That's the key. I'm not too sure POVRay's radiosity gives an unbiased 
result even with all settings turned up to the best quality. If there is 
just one reflective surface in the mix I'm downright sure it does not...

Even throwing photons in I believe there are situations POV will not 
handle correctly. Can we do a wall lighting one side of a room because 
it is itself lit by a light reflecting in a mirror?

Yeah, I should try myself :-)

-- 
Vincent

Post a reply to this message

> Orchid XP v7 wrote:
>> As best I can tell, the algorithm described just sounds like POV-Ray's 
>> radiosity with an infinitely low error_bound. (I.e., always resample.) 
>> But applied to *all* terms, not just diffuse illumination...
>>
> 
> That's the key. I'm not too sure POVRay's radiosity gives an unbiased 
> result even with all settings turned up to the best quality. If there is 
> just one reflective surface in the mix I'm downright sure it does not...
> 
> Even throwing photons in I believe there are situations POV will not 
> handle correctly. Can we do a wall lighting one side of a room because 
> it is itself lit by a light reflecting in a mirror?

"Light shines onto a white diffuse wall. Reflected light from the wall 
creates caustics through a glass sphere. That is diffuse -> Specular 
(caustics). I got the idea for this scene from a document by Henrik Wann 
Jensen describing some difficult test scenes."

http://www.winosi.onlinehome.de/Gallery_t15.htm

Post a reply to this message

Nicolas Alvarez wrote:

>> Even throwing photons in I believe there are situations POV will not 
>> handle correctly. Can we do a wall lighting one side of a room because 
>> it is itself lit by a light reflecting in a mirror?
> 
> "Light shines onto a white diffuse wall. Reflected light from the wall 
> creates caustics through a glass sphere. That is diffuse -> Specular 
> (caustics). I got the idea for this scene from a document by Henrik Wann 
> Jensen describing some difficult test scenes."
> 
> http://www.winosi.onlinehome.de/Gallery_t15.htm

Ah yes now I remember seeing this one... Well it is expected that 
unbiased renderer are able to tackle this, of course...

-- 
Vincent

Post a reply to this message

Vincent Le Chevalier wrote:

> That's the key. I'm not too sure POVRay's radiosity gives an unbiased 
> result even with all settings turned up to the best quality. If there is 
> just one reflective surface in the mix I'm downright sure it does not...

I do recall somebody rendered a radiosity scene with no photon maps at 
all, and yet a large curved mirror still produced correct caustics. (It 
was a damn long time ago now though, so I'd never be able to find the 
image...)

> Even throwing photons in I believe there are situations POV will not 
> handle correctly. Can we do a wall lighting one side of a room because 
> it is itself lit by a light reflecting in a mirror?
> 
> Yeah, I should try myself :-)

Radiosity ought to do it - although it may or may not require some 
fairly crazy settings...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v7 wrote:
> As best I can tell, the algorithm described just sounds like POV-Ray's 
> radiosity with an infinitely low error_bound. (I.e., always resample.) 

don't forget higher than 1600 count limit...

Post a reply to this message

Orchid XP v7 <voi### [at] devnull> wrote:
> As best I can tell, the algorithm described just sounds like POV-Ray's 
> radiosity with an infinitely low error_bound. (I.e., always resample.) 
> But applied to *all* terms, not just diffuse illumination...

  Except that POV-Ray doesn't have the concept of BRDFs.

  Also, brute-force renderers take a slightly different approach at
rendering than POV-Ray's radiosity. What the latter does is that when
a ray intersects a surface, it sends tons of rays to all directions
(unless it can interpolate from nearby values calculated earlier).
AFAIK brute-force renderers only send one ray to a random direction.
This means that a ray hitting a surface doesn't spawn a thousand new
rays, but only one.

  There are advantages and disadvantages with this. The advantage is that
the renderer never gets stuck on things like highly reflective/refractive
objects because it's only following one path, not thousands (each one of
which in turn would spawn thousands of new paths, etc). Another advantage
is that you don't have to fix a number of rays to send per intersection
point, but you can simply let it run for as long as you want.

  The disadvantage is the graininess, of course. Overall, however, the
correct result is approached faster in this way than with the method
POV-Ray uses. (This is assuming we are looking for the perfect solution.)

-- 
                                                          - Warp

Post a reply to this message

Warp escribió:
>   There are advantages and disadvantages with this. The advantage is that
> the renderer never gets stuck on things like highly reflective/refractive
> objects because it's only following one path, not thousands (each one of
> which in turn would spawn thousands of new paths, etc). Another advantage
> is that you don't have to fix a number of rays to send per intersection
> point, but you can simply let it run for as long as you want.
> 
>   The disadvantage is the graininess, of course. Overall, however, the
> correct result is approached faster in this way than with the method
> POV-Ray uses. (This is assuming we are looking for the perfect solution.)
> 

I think another advantage is it works well with deadlines. Instead of 
worrying if you'll finish the render by the IRTC deadline, you just let 
it run till the last minute. It may be more or less grainy, but at least 
you have the whole image. With a normal raytracer, you'd have the top 
half of the image...

Post a reply to this message

Orchid XP v7 wrote:
> Vincent Le Chevalier wrote:
> 
>> That's the key. I'm not too sure POVRay's radiosity gives an unbiased 
>> result even with all settings turned up to the best quality. If there 
>> is just one reflective surface in the mix I'm downright sure it does 
>> not...
> 
> I do recall somebody rendered a radiosity scene with no photon maps at 
> all, and yet a large curved mirror still produced correct caustics. (It 
> was a damn long time ago now though, so I'd never be able to find the 
> image...)
> 

Well, strange...
In the docs it is written:
"
The brightness of radiosity in POV-Ray is based on two things:

    1. the amount of light "gathered"
    2. the 'diffuse' property of the surface finish
"
<http://www.povray.org/documentation/view/3.6.1/270/>

So it seems pretty clear that radiosity does not take the specular path, 
ever, because the "reflection" property is not considered. Which means 
that rays doing:
Light->specular->diffuse->diffuse->camera (the situation I described)
or Light->diffuse->specular->diffuse->camera (as in the image Nicolas 
posted)
are never created.

But if there is a scene proving me wrong I'd be glad to see it.

-- 
Vincent

Post a reply to this message

On Thu, 21 Feb 2008 17:12:04 -0300, nemesis wrote:

> Jim Henderson wrote:
>> Ah, so you've decided Darren and I were right after all, then? ;-)
> 
> yes, now I'm a satanist. :P </sarcasm>

Ah, then you didn't understand our point. ;-)

Jim

Post a reply to this message