[Radiance-general] Mkillum usage? with stained glass....

[Rob Guglielmetti]

On Mar 15, 2004, at 8:05 PM, John Sutherland wrote:
> Firstly – Ive read a lot about the mkillum approach, so for stained 
> glass I’m guessing I create an object that is textured glass material, 
> and have the mkillum secondary source behind it? Shining through. This 
> presumably will project the colours of the glass into the room?

I still struggle with the correct usage of illum, but let's have a go 
at it.  Generally, for most accurate results, the glass panes 
themselves would be converted to illums.  But in this case, I'm 
imagining that your window is comprised of lots of irregularly shaped 
pieces of glass.  These will never resolve properly when mkillum tries 
to convert these polygons into illums.  I'd guess your best bet is to 
create some impostor geometry just on the *inside* of the stained glass 
window geometry.

> Secondly – You can see mkillum sources from the rear side, I’m 
> guessing you turn them off for renders of the outside of buildings 
> looking at the windows?

Hmm.  For an exterior rendering, I'd turn them off, but not for the 
reason you mention.  Illums are invisible when visible.  That is to 
say, in views where the illum geometry is directly visible, it is not 
rendered.  So, don't worry about excluding illums from your scenes if 
visibility is ever a concern.

But in this example case, a rendering outside the building, the illums 
aren't contributing anything to the image so why even waste time adding 
them to the scene and calculating them?  So, for your exterior views 
you should use a different set of parameters, and leave out the stained 
window illum geometry (the stained glass itself will get illuminated by 
your sky & sun and thus will still look correct.)

> Secondly – I’m really interested in the internal lighting, how bright 
> the inside will be at different times of day etc. When I render 
> internally with no av settings (rview), and set e to 1 is it bringing 
> the environment into a range that we can see? What im saying is, how 
> do I know what settings to use to get the most realistic 
> representation of what it would have been like to the human eye? I 
> know this is a tricky problem considering that the human eye adjusts 
> to light. How can I approach this problem?

Whoo-whee.  Well, that is a monster question.  Simple answer is that 
there is no simple answer to this.  Every model you do in Radiance will 
present new challenges, but Radiance by and large can handle whatever 
you throw at it.  Have you used the rad program yet?  Rad will 
certainly get you going in the right direction quickly.  You can also 
save the various parameters that rad feeds to rpict and inspect them 
later.  It's quite educational.  You're on the right track with illum 
there, that will improve your results and accuracy immensely.  Careful 
application of mkillum, a handful of ambient bounces and say Q=M V=H 
D=[M or H] for your rad settings and you'll be in good shape, in 
general.

Your question about eye adaptation leads me to one of my favorite tools 
in the Radiance suite, pcond.  Pcond is a tonemapping program which 
already has a very useful commandline parameter (-h) built into it to 
simulate human visibility in a Radiance image.  So, after you generate 
your radiance images, you simply run them through pcond like so:

pcond -h image.pic image-human.pic

Pcond takes the radiance image and performs a tonemapping operation.  
The results are fantastic, and it's really one of the best ways to 
demonstrate the power of Radiance to people.

Have fun, and welcome aboard!

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