estimating -av values, color

Hi list,

I have a question regarding the so-often given advice to use a neutral -av setting. Why? Is it assumed that in general, the light conditions in a building are like that? I have used a -av of .5 .5 .5 here, together with -ae excluded objects (I wrote enough questions on that topic ;-). This results in a strange image, as the calculated ambient is redish, and the ambient-excluded objects appear very strange (like blueish) because they are the only objects illuminated by this neutral color. I can understand that if using -ab 0, setting a colored -av would not make sense as ALL indirect illumination would be neutral than.

So, am I right that in the case here (I use the -av settings only for the -ae excluded objects in a scene that is rendered with -ab 2), using a neutral value would be simply wrong and I have to use the "measured" (from rview) radiance?

TIA+CU Lars.

Hi Lars,

Estimating ambient values is part science and part art as far as I have learned over the years. I don't think that I know of a situation where we have ever used a non neutral ambient value. I suppose if one were trying to more accurately account for certain kind of atmospheric shifts this might be a case to do this....?

Without knowing what your scene looks like (pictures help) it is difficult to help diagnose what is going on. My suscpicion though is that an ambient value of .5 .5 .5 may be too high for the given scene, thus the odd looking ("glowing") results. A very good option would be to use the "compamb" script that Greg wrote. This runs a pass on a small image using fairly high parameters and then does some calculations to come up with a good ambient value for the scene. This is pretty easy to use and does not take that long to return the results.

Note the av value affects the whole simulation not just the excluded objects.

-Jack

Lars O. Grobe wrote:

···

Hi list,

I have a question regarding the so-often given advice to use a neutral -av setting. Why? Is it assumed that in general, the light conditions in a building are like that? I have used a -av of .5 .5 .5 here, together with -ae excluded objects (I wrote enough questions on that topic ;-). This results in a strange image, as the calculated ambient is redish, and the ambient-excluded objects appear very strange (like blueish) because they are the only objects illuminated by this neutral color. I can understand that if using -ab 0, setting a colored -av would not make sense as ALL indirect illumination would be neutral than.

So, am I right that in the case here (I use the -av settings only for the -ae excluded objects in a scene that is rendered with -ab 2), using a neutral value would be simply wrong and I have to use the "measured" (from rview) radiance?

TIA+CU Lars.

------------------------------------------------------------------------

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

--
# John E. de Valpine
# president
#
# visarc incorporated
# http://www.visarc.com
#
# channeling technology for superior design and construction

Hi Lars,

you can think of the av value being a substitution for the rest amount of light hovering in the room which gets ommitted in your calculations when you follow
ambient rays only a certain number of bounces deep, instead of following them infinitely deep. And if you have a neutral source, but red walls, the diffuse indirect light will be red, and so should also be the av setting, to be a correct compensation...

So generally your're right in considering the color bias of the scene also for the constant ambient approximation. But definitely also keep in mind Jacks remark on -av affecting all objects...

-cb

So generally your're right in considering the color bias of the scene also for the constant ambient approximation. But definitely also keep in mind Jacks remark on -av affecting all objects...

Hi,

thanks to Jack and Carsten for the responses, this list is always of great help. I didn't really use -av in previous projects, but need it now because the complexity of the scene forces me to exclude modifiers from the ambient calculation.

As far as I understood, the influence of the -av setting (on objects not excluded from the ambient calculation) can be dimmed by the -aw setting, and as I start my renderings with an "ouverture calculation" the computed ambient distribution should be stable enough. Am I wrong here?

CU Lars.

As far as I understood, the influence of the -av setting (on objects not excluded from the ambient calculation) can be dimmed by the -aw setting, and as I start my renderings with an "ouverture calculation" the computed ambient distribution should be stable enough. Am I wrong here?

hmm, I never used -aw so far, prob. I was to lazy to count the ambient vals to find out which number to set for -aw...

I think one point is important to consider: setting a constant 'rest' ambient value -av for the whole scene is a very crude thing which can be applied moderately to prevent certain regions appearing unnaturally dark, but it will be difficult to use -av as some kind of substitution on objects otherwise excluded from ambient ray tracing..
It quickly looks strange an unnatural either. I remember that it even was difficult to tweak things that way in POVRay, where you could (still can?) set
an ambient value for each material (resp. finish) separately..

-cb

PS -btw, the general problem was one of the backgrounds why I implemented such hacks as the secondary ambient material in radzilla for objects with complicated patterns but still some sort of base color, or the 'scene' with its possibility of setting different ambient parameters, etc.
All serve in extending the range in which the raytracing ambient calc can still be used without letting rendering times explode. Because in the end the ambient rays are the only really working way of producing nice results, all other workarounds will sooner or later reach their limits..)

Hey, now that would be cool. Is there any way that could be added to Radiance? The ability to have multiple -av's for various materials would increase the usability of -ae, I think. Especially for large scenes that incorporate interior & exterior elements...

···

On Jul 23, 2005, at 4:53 AM, Carsten Bauer wrote:

I think one point is important to consider: setting a constant 'rest' ambient value -av for the whole scene is a very crude thing which can be applied moderately to prevent certain regions appearing unnaturally dark, but it will be difficult to use -av as some kind of substitution on objects otherwise excluded from ambient ray tracing..
It quickly looks strange an unnatural either. I remember that it even was difficult to tweak things that way in POVRay, where you could (still can?) set
an ambient value for each material (resp. finish) separately..

=================
    Rob Guglielmetti
www.rumblestrip.org

Rob Guglielmetti wrote:

Hey, now that would be cool. Is there any way that could be added to Radiance? The ability to have multiple -av's for various materials would increase the usability of -ae, I think. Especially for large scenes that incorporate interior & exterior elements..

Hi Rob,
implementing this is more or less rather easy, I remember having played around with it some years ago already, but I didn't find it rewarding resp. worth the effort. The basic problem is always the 'flat' illumination caused by the constant value.

When it comes to image generation, if you have large scenes with exterior and interior parts both of importance for display in the image, and/or e.g complex stuff like trees etc, you should give the 'scene' feature of radzilla a try, with it's possibility of setting a separate parameter set meaning also separate caches.
It depends of course on the type of object. Computers are fast now, so even in million polygon octrees like a tree e.g a crude ambient cache with -ab1 and -ad 60 or 100 or so is trackable. In such an obect the errors caused by the crude cache are of no importance, and it definitely gives better results than any constant value approximation.
In other cases, if the outside part is of lesser importance but still somehow visible in the image, setting -ab 0 and a constant av for the whole outside part might of course suffice, and it can save you the time and work of pasting some background stuff into the image with photoshop etc later..

-cb

Hi Rob,
implementing this is more or less rather easy, I remember having played
around with it some years ago already, but I didn't find it rewarding
resp. worth the effort. The basic problem is always the 'flat'
illumination caused by the constant value.

Agreed, but the direct calculation still puts some variation on those
objects that have the constant -av approximation, and if you could have a
variety of -av approximations to apply, one could have a fair amount of
variation in the scene, even with the excludes. It could help, anyway.

When it comes to image generation, if you have large scenes with
exterior and interior parts both of importance for display in the image,
and/or e.g complex stuff like trees etc, you should give the 'scene'
feature of radzilla a try, with it's possibility of setting a separate
parameter set meaning also separate caches.

I really, really should. I recall reading about that and thinking that
that was a very good option for these kinds of scenes. (I also think that
your falsecolor material is just too cool.)

It depends of course on the type of object. Computers are fast now, so
even in million polygon octrees like a tree e.g a crude ambient cache
with -ab1 and -ad 60 or 100 or so is trackable. In such an obect the
errors caused by the crude cache are of no importance, and it
definitely gives better results than any constant value approximation.
In other cases, if the outside part is of lesser importance but still
somehow visible in the image, setting -ab 0 and a constant av for the
whole outside part might of course suffice, and it can save you the time
and work of pasting some background stuff into the image with photoshop
etc later..

Photoshop has no place in my Radiance workflow. =8-)

···

On Mon, July 25, 2005 2:06 pm, Carsten Bauer said:

--
Rob Guglielmetti
www.rumblestrip.org

Rob Guglielmetti wrote:

variety of -av approximations to apply, one could have a fair amount of
variation in the scene, even with the excludes. It could help, anyway.

well, the easiest way of implementing it is an augmented material type with possibility of specifying an additional ambient value, means addtional to everything, also to an existing av.. If you want to get a feeling how this might look in principle then you can play around a bit the with 'ghost' material in current radzilla, which has this feature. Although here the original reason for offering it is different, I thought of special cases where one wants to set markers in the scene, or for some purely artistic stuff.

One might give it a try and offer this opportunity, although personally I still think it will be rather confusing (cf. my remarks on the POVRay case..)

Messing with the ambient caches's av itself is btw problematic, because of the recursive nature of the ambient calc.

Photoshop has no place in my Radiance workflow. =8-)

well, sorry, the honesty, of course your're right. How could I.... :slight_smile:
BTW, evidently Photoshop has no place in my rendering workflow either.
I use Gimp ...

-cb