Hello. I am trying to separate the direct and indirect
specular and indirect diffuse calculation in Radiance.
Actually, I am trying to duplicate the "artificial
separation" provided in Radiance book, plate 23.
I got a little confused with the rendering parameters:
Lets say I have an image, which has all
direct + indirect specular + indirect diffuse
computation, with parameters
-av 0.5 0.5 0.5
-ab 2
-st 0.001
INDIRECT=1
Would it be sufficient to change
-av 0 0 0
-ab 0
-st 1
INDIRECT=0
so that I can get an image with just the direct
computations? Or do I have to change other parameters?
Accordingly, if change the ?st parameter back to
0.001; would it be give direct + indirect specular
component?
Thanks.
Ada
···
__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com
Hi Ada,
I think there's no way to do this correctly without understanding and hacking the source code in src/rt/normal.c. That's what I had to do to generate this image in the first place. If you just want a copy of the image, I would be happy to provide one. You won't be able to get this result by altering the command line options or the material parameters alone. If your specular surface(s) has zero roughness, you might be able to get close by rendering pictures of your scene with and without the specular component(s), and subracting the resulting pictures. As for separating the interreflected component, the same picture subtraction should work between one rendered with -ab 1 -av [whatever] and -ab 0 -av 0 0 0.
To subtract two images:
pcomb img1.pic -s -1 img2.pic > img1-img2.pic
The images must be rendered at the same size and resolution, setting the rpict parameters -pj 0 and -ps 1. You can pass the result through pfilt to improve the final appearance, but I don't recommend doing it before.
I hope this helps.
-Greg
···
Ada Fraser <[email protected]> wrote:
Hello. I am trying to separate the direct and indirect
specular and indirect diffuse calculation in Radiance.
Actually, I am trying to duplicate the "artificial
separation" provided in Radiance book, plate 23.
I got a little confused with the rendering parameters:
Lets say I have an image, which has all
direct + indirect specular + indirect diffuse
computation, with parameters
-av 0.5 0.5 0.5
-ab 2
-st 0.001
INDIRECT=1
Would it be sufficient to change
-av 0 0 0
-ab 0
-st 1
INDIRECT=0
so that I can get an image with just the direct
computations? Or do I have to change other parameters?
Accordingly, if change the ?st parameter back to
0.001; would it be give direct + indirect specular
component?
Thanks.
Ada
Hello. Thank you for the answer Greg. It is very
helpful. I have one follow up question: When I set the
rendering parameters -ab 0 -av 0 0 0, I am eliminating
the interreflected component. But I still see some
reflection of the interior surfaces on my window
glazing. Since there is no interreflection, why is
this happening?
Thanks again.
Ada
···
--- Greg Ward <[email protected]> wrote:
Hi Ada,
I think there's no way to do this correctly without
understanding and
hacking the source code in src/rt/normal.c. That's
what I had to do to
generate this image in the first place. If you just
want a copy of the
image, I would be happy to provide one. You won't
be able to get this
result by altering the command line options or the
material parameters
alone. If your specular surface(s) has zero
roughness, you might be
able to get close by rendering pictures of your
scene with and without
the specular component(s), and subracting the
resulting pictures. As
for separating the interreflected component, the
same picture
subtraction should work between one rendered with
-ab 1 -av [whatever]
and -ab 0 -av 0 0 0.
To subtract two images:
pcomb img1.pic -s -1 img2.pic > img1-img2.pic
The images must be rendered at the same size and
resolution, setting
the rpict parameters -pj 0 and -ps 1. You can pass
the result through
pfilt to improve the final appearance, but I don't
recommend doing it
before.
I hope this helps.
-Greg
Ada Fraser <[email protected]> wrote:
>
> Hello. I am trying to separate the direct and
indirect
> specular and indirect diffuse calculation in
Radiance.
> Actually, I am trying to duplicate the "artificial
> separation" provided in Radiance book, plate 23.
>
> I got a little confused with the rendering
parameters:
> Lets say I have an image, which has all
> direct + indirect specular + indirect diffuse
> computation, with parameters
> -av 0.5 0.5 0.5
> -ab 2
> -st 0.001
> INDIRECT=1
>
> Would it be sufficient to change
> -av 0 0 0
> -ab 0
> -st 1
> INDIRECT=0
> so that I can get an image with just the direct
> computations? Or do I have to change other
parameters?
>
>
> Accordingly, if change the ?st parameter back to
> 0.001; would it be give direct + indirect specular
> component?
>
> Thanks.
> Ada
__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com
Hi Ada,
I should have been more specific -- setting -ab 0 -av 0 0 0 eliminates _diffuse_ interreflections in the calculation -- specular reflections will still be present. The only way to get rid of specular reflections (reliably) is to eliminate specular components from your materials. In the case of a glass surface, you would have to substitute trans with a zero reflected component.
-Greg
Ada Fraser <[email protected]> writes:
···
Hello. Thank you for the answer Greg. It is very
helpful. I have one follow up question: When I set the
rendering parameters -ab 0 -av 0 0 0, I am eliminating
the interreflected component. But I still see some
reflection of the interior surfaces on my window
glazing. Since there is no interreflection, why is
this happening?
Thanks again.
Ada
Greg Ward wrote:
Hi Ada,
I should have been more specific -- setting -ab 0 -av 0 0 0 eliminates
_diffuse_ interreflections in the calculation -- specular reflections
will still be present. The only way to get rid of specular reflections
(reliably) is to eliminate specular components from your materials. In
the case of a glass surface, you would have to substitute trans with a
zero reflected component.
Doesn't -lr 0 truncate *all* interreflection, specular and diffuse?
--Roland
···
--
END OF LINE. (MCP)
From: Roland Schregle <[email protected]>
Greg Ward wrote:
Hi Ada,
I should have been more specific -- setting -ab 0 -av 0 0 0 eliminates
_diffuse_ interreflections in the calculation -- specular reflections
will still be present. The only way to get rid of specular reflections
(reliably) is to eliminate specular components from your materials. In
the case of a glass surface, you would have to substitute trans with a
zero reflected component.
Doesn't -lr 0 truncate *all* interreflection, specular and diffuse?
--Roland
Well... It's never simple with Radiance, is it? You would think that -lr truncates reflections, and usually it does, but you'll still get the direct component in there, which includes the specular highlight if the roughness value is non-zero.
-Greg
Greg Ward wrote:
> From: Roland Schregle <[email protected]>
>
> Greg Ward wrote:
>>
>> Hi Ada,
>>
>> I should have been more specific -- setting -ab 0 -av 0 0 0 eliminates
>> _diffuse_ interreflections in the calculation -- specular reflections
>> will still be present. The only way to get rid of specular
>> reflections
>> (reliably) is to eliminate specular components from your materials.
>> In
>> the case of a glass surface, you would have to substitute trans with a
>> zero reflected component.
>
> Doesn't -lr 0 truncate *all* interreflection, specular and diffuse?
>
> --Roland
>
Well... It's never simple with Radiance, is it? You would think that
-lr truncates reflections, and usually it does, but you'll still get
the direct component in there, which includes the specular highlight if
the roughness value is non-zero.
Yeah, but that's not specular INTERreflection, which Ada was trying to
suppress. Or not? What was the original question? :^)
And just to make it a little more complicated: how does this affect
virtual sources?
--Confused
···
--
END OF LINE. (MCP)
I was wondering if I could get some input from the experienced Radiance users out there regarding rendering parameters. I am currently rendering an extremely basic room (13.5'x13.5'x14') with one indirect light fixture. However, accuracy is extremely important. So I am in the process of running test runs to see what would be the optimal rendering parameters within the time limit I have.
I had been getting blotchy results, so I have been doubling the -ad parameter. My most recent rendering parameters were:
-i -t 600 -dt 0.04 -dc 1 -ds 0.01 -dj 0.6 -dr 3 -dp 4096 -st 0.01 -sj 1 -av 0 0 0 -aw 0 -ab 3 -ad 4096 -as 2048 -ar 256 -aE ../ambient_exclude -ms 2.7 -lr 12 -lw 0.0005 -aa 0.04 -af b_b_m.amb -ps 1 -pt 0.04 -x 512 -y 512 -pa 0
I did an overture rendering at -x 64 -y 64 prior to creating the image at 512x512.
I am, still getting what I feel to be significant blotchiness. The image can be seen at:
http:/homepage.mac.com/jsan
So, my questions are:
Are there any immediate red-flags you see in my rendering parameters that seem incorrect?
Do I simply need to keep increasing the -ad to get better results, or would lowering the -pt help both the visual and calculation accuracy?
If I pfilt the image with -x /2 -y /2, will that merge the values of the pixels enough to make the values at each pixel relatively accurate?
Just as background, this image took about 12 hours to render (with the overture calculations) using rpiece on a dual G5 Macintosh with 1.5 GB of RAM and a 2.24 P4 with 1 GB of RAM. I have about 200 similar situations to run (more light fixtures, larger room, etc.) So increasing the -ad to 8192 seems to be prohibitively expensive at this point.
Thanks.
Sorry, forgot to mention that the view taken of the image is a plan view.
···
On Jan 4, 2004, at 12:25 PM, John S. An wrote:
I was wondering if I could get some input from the experienced Radiance users out there regarding rendering parameters. I am currently rendering an extremely basic room (13.5'x13.5'x14') with one indirect light fixture. However, accuracy is extremely important. So I am in the process of running test runs to see what would be the optimal rendering parameters within the time limit I have.
I had been getting blotchy results, so I have been doubling the -ad parameter. My most recent rendering parameters were:
-i -t 600 -dt 0.04 -dc 1 -ds 0.01 -dj 0.6 -dr 3 -dp 4096 -st 0.01 -sj 1 -av 0 0 0 -aw 0 -ab 3 -ad 4096 -as 2048 -ar 256 -aE ../ambient_exclude -ms 2.7 -lr 12 -lw 0.0005 -aa 0.04 -af b_b_m.amb -ps 1 -pt 0.04 -x 512 -y 512 -pa 0
I did an overture rendering at -x 64 -y 64 prior to creating the image at 512x512.
I am, still getting what I feel to be significant blotchiness. The image can be seen at:
http:/homepage.mac.com/jsan
So, my questions are:
Are there any immediate red-flags you see in my rendering parameters that seem incorrect?
Do I simply need to keep increasing the -ad to get better results, or would lowering the -pt help both the visual and calculation accuracy?
If I pfilt the image with -x /2 -y /2, will that merge the values of the pixels enough to make the values at each pixel relatively accurate?
Just as background, this image took about 12 hours to render (with the overture calculations) using rpiece on a dual G5 Macintosh with 1.5 GB of RAM and a 2.24 P4 with 1 GB of RAM. I have about 200 similar situations to run (more light fixtures, larger room, etc.) So increasing the -ad to 8192 seems to be prohibitively expensive at this point.
Thanks.
Regarding the problems I am having with simple room rendering, I have zipped and placed all the .rad, .mat, and control files at the same address (http://homepage.mac.com/jsan).
Also, I finished rendering another image of a room twice the size of the original (13'-6 x 27') and with 18 bare bulbs. This time, there is a weird moir�-like effect in the rendering. All of the parameters were, for the most part, identical to the previous case (the only changes were that I upped the -y parameter to 1024 to account for the larger room).
Any ideas as to what is going wrong?
John An
Hi John,
I don't really have time right now to look at your input files, but it seems that your problems are associated with the light source sampling. Specifically, the original image with the single fixture seems to have some kind of geometry in front of it, which is going to cause randomness in the shadows produced. You should either get rid of this bit of geometry (looks like three vanes of a fan) or use an illum without such obstructions in front of it.
In the second case where you have many sources, what you are seeing is the result of the direct jitter. If you set -dj 0 these artifacts should disappear.
-Greg
···
From: "John S. An" <[email protected]>
Date: January 5, 2004 2:40:33 PM PST
Regarding the problems I am having with simple room rendering, I have zipped and placed all the .rad, .mat, and control files at the same address (http://homepage.mac.com/jsan).
Also, I finished rendering another image of a room twice the size of the original (13'-6 x 27') and with 18 bare bulbs. This time, there is a weird moir�-like effect in the rendering. All of the parameters were, for the most part, identical to the previous case (the only changes were that I upped the -y parameter to 1024 to account for the larger room).
Any ideas as to what is going wrong?
John An
Greg,
That did the trick. The geometry above the luminaire was the cable which would suspend it. I simply removed it, and all seems fine now. I feel so stupid, I spent an entire week increasing the -ad parameter and not getting anywhere.
For my purposes, it is ok to remove the cables at this point from the model. However, for suspended light fixtures, it doesn't seem possible to ever create an illum that would encompass everything including the cables without having the illum sphere touch up against the ceiling. This would be more for indirect fixtures than for direct fixtures. If I needed to keep the suspension cables in the model, what would be the most proper way of defining the light source?
John
···
On Jan 6, 2004, at 3:12 PM, Greg Ward wrote:
Hi John,
I don't really have time right now to look at your input files, but it seems that your problems are associated with the light source sampling. Specifically, the original image with the single fixture seems to have some kind of geometry in front of it, which is going to cause randomness in the shadows produced. You should either get rid of this bit of geometry (looks like three vanes of a fan) or use an illum without such obstructions in front of it.
In the second case where you have many sources, what you are seeing is the result of the direct jitter. If you set -dj 0 these artifacts should disappear.
-Greg
From: "John S. An" <[email protected]>
Date: January 5, 2004 2:40:33 PM PST
Regarding the problems I am having with simple room rendering, I have zipped and placed all the .rad, .mat, and control files at the same address (http://homepage.mac.com/jsan).
Also, I finished rendering another image of a room twice the size of the original (13'-6 x 27') and with 18 bare bulbs. This time, there is a weird moir�-like effect in the rendering. All of the parameters were, for the most part, identical to the previous case (the only changes were that I upped the -y parameter to 1024 to account for the larger room).
Any ideas as to what is going wrong?
John An
Hi John,
I admit to being still a little puzzled by your results. I'd really have to study the input files and run some renderings to figure out what's going on with them. Normally, a little bit of geometry in front of the light source shouldn't cause splotches, especially if the -dj option is used.
-G
···
From: "John S. An" <[email protected]>
Date: January 8, 2004 12:29:32 PM PST
Greg,
That did the trick. The geometry above the luminaire was the cable which would suspend it. I simply removed it, and all seems fine now. I feel so stupid, I spent an entire week increasing the -ad parameter and not getting anywhere.
For my purposes, it is ok to remove the cables at this point from the model. However, for suspended light fixtures, it doesn't seem possible to ever create an illum that would encompass everything including the cables without having the illum sphere touch up against the ceiling. This would be more for indirect fixtures than for direct fixtures. If I needed to keep the suspension cables in the model, what would be the most proper way of defining the light source?
John