I figured I would post this in General first, to test the waters.

Povray appears to have a problem with radiosity and negative surface colors.
I have a material which has a solid pigment of {rgb<-1,-1,0>}, and usually
it works quite nicely and produces nice deep blue colors in its
interreflections with other objects. Occasionally, though, I get mysterious
artifacts:
http://idt.gatech.edu/~cashmore/pov/debug01a.jpg

That black shadow that looks like it's being produced by a sphere is in
fact, not the shadow of any object that exists in the scene.

This is exacerbated by the use of reflections. In the following I am
alternating the black spheres with reflective ones. The artifacts should
clearly be incorrect, since the two images have different error_bounds
http://idt.gatech.edu/~cashmore/pov/debug01_05.jpg error_bound 0.5
http://idt.gatech.edu/~cashmore/pov/debug01_02.jpg error_bound 0.2

If I take away the negative color, the problem goes away, but I'd prefer to
not need to do that. I suspect the problem lies with the implementation of
adc_bailout, in that pov sees the net magnitude of the reflected color as
being less than zero, and short circuts to make the resulting color darker
than it would be even otherwise.

Any thoughts or comments?
Can people reproduce this on other architectures?

Here is my source:
http://idt.gatech.edu/~cashmore/pov/debug01.pov

Post a reply to this message

Calvin wrote:
> I figured I would post this in General first, to test the waters.
> 
> Povray appears to have a problem with radiosity and negative surface colors.
> I have a material which has a solid pigment of {rgb<-1,-1,0>}, and usually
> it works quite nicely and produces nice deep blue colors in its
> interreflections with other objects. Occasionally, though, I get mysterious
> artifacts:
> http://idt.gatech.edu/~cashmore/pov/debug01a.jpg
> 
> That black shadow that looks like it's being produced by a sphere is in
> fact, not the shadow of any object that exists in the scene.
> 
> This is exacerbated by the use of reflections. In the following I am
> alternating the black spheres with reflective ones. The artifacts should
> clearly be incorrect, since the two images have different error_bounds
> http://idt.gatech.edu/~cashmore/pov/debug01_05.jpg error_bound 0.5
> http://idt.gatech.edu/~cashmore/pov/debug01_02.jpg error_bound 0.2
> 
> If I take away the negative color, the problem goes away, but I'd prefer to
> not need to do that. I suspect the problem lies with the implementation of
> adc_bailout, in that pov sees the net magnitude of the reflected color as
> being less than zero, and short circuts to make the resulting color darker
> than it would be even otherwise.
> 
> Any thoughts or comments?
> Can people reproduce this on other architectures?
Other than what?
I tried it on a W2k machine with POV 3.6.
I also get your artifacts. Artifacts are not in the same place
as in your picture. That is probably OK, as every time I render
it, the artifacts are in different places. If I change the
1*<-1,-1,0> to 1+<-1,-1,0> the scene seems to be OK.

> 
> Here is my source:
> http://idt.gatech.edu/~cashmore/pov/debug01.pov
> 
> 
>

Post a reply to this message

> > Any thoughts or comments?
> > Can people reproduce this on other architectures?
> Other than what?
> I tried it on a W2k machine with POV 3.6.
> I also get your artifacts. Artifacts are not in the same place
> as in your picture. That is probably OK, as every time I render
> it, the artifacts are in different places. If I change the
> 1*<-1,-1,0> to 1+<-1,-1,0> the scene seems to be OK.

I'm sorry, I neglected to mention that-
I'm on Windows XP with POV 3.6 as well.

I actually am not getting changes in the artifacts, they remain consistent
for each render. If they are random for you, though, I suspect this might
be genuine buggy behavior.

Post a reply to this message

Hmm, I tried it too.  Re-running at the same resolution on the same machine
seems to give the same results, but changing either the machine or the
resolution changes the image.
Charles

Post a reply to this message

> Any thoughts or comments?

There's no such thing as negative colors.
How do you expect "minus blue" to look?
Should it be red? Or green? Or just black?

Post a reply to this message

Tim Attwood wrote:
>>Any thoughts or comments?
> 
> 
> There's no such thing as negative colors.
> How do you expect "minus blue" to look?
> Should it be red? Or green? Or just black?
> 
> 
I would have thought yellow.

Post a reply to this message

>> There's no such thing as negative colors.
>> How do you expect "minus blue" to look?
> I would have thought yellow.

rgb <0, 0, -1> has no components that would contribute to yellowness. 
Anything that might emit "yellow" light is zero. The closest might be 
that it "emits" negative blue, making things nearby more yellow, sucking 
away the blue via radiosity.

Or maybe it's a quark, where it would be a red/antiblue quark or 
something. ;-)

-- 
   Darren New / San Diego, CA, USA (PST)
     Scruffitarianism - Where T-shirt, jeans,
     and a three-day beard are "Sunday Best."

Post a reply to this message

Darren New nous apporta ses lumieres en ce 05-01-2007 11:56:
>>> There's no such thing as negative colors.
>>> How do you expect "minus blue" to look?
>> I would have thought yellow.

> rgb <0, 0, -1> has no components that would contribute to yellowness. 
> Anything that might emit "yellow" light is zero. The closest might be 
> that it "emits" negative blue, making things nearby more yellow, sucking 
> away the blue via radiosity.

> Or maybe it's a quark, where it would be a red/antiblue quark or 
> something. ;-)

In a radiosity scene, a negative colour will effectively suck out the colours 
from it's environment.
You also can have negative light that will darken your scene and any high 
ambient object.
Also, a negative colour will apears differently in reflections and refractions 
than plain black.

-- 
Alain
-------------------------------------------------
Fine day to work off excess energy.  Steal something heavy.

Post a reply to this message

>
> > rgb <0, 0, -1> has no components that would contribute to yellowness.
> > Anything that might emit "yellow" light is zero. The closest might be
> > that it "emits" negative blue, making things nearby more yellow, sucking
> > away the blue via radiosity.
>
> In a radiosity scene, a negative colour will effectively suck out the colours
> from it's environment.
> You also can have negative light that will darken your scene and any high
> ambient object.
> Also, a negative colour will apears differently in reflections and refractions
> than plain black.
>

In my scene, it's actually negative yellow, so it's producing blue
coloration. Negative light isn't natural, but has been supported since the
early days of POV. Really the goal is not to produce direct coloration, but
to suck the color out of other objects. Negative light sources can be done
to produce similar results, effectively producing "extra dark" parts of the
scene. This can be used to fake shadows, but I happen to like the sort of
unnatural effect that it can give.

I'll link to a couple of images in a short time that shows some of these
effects at work.

Post a reply to this message