Hello,
after ~year of beeing too bussy to use [seriously] PovRAY,
I have few questions now when I try to go back to using it :)

1) blurred reflection - I see it is suggested (in FAQ) to use several normal
patterns (or textures) and strong antialias.
How about using the old method - because I think it might benefit in speed,
by using adaptive number of samples (same as in camera focal blur). The AA
of entire image is not as good probably 

2) radiosity and light fading - natural light behaves like with fade_power 2
right? What about the radiosity? It is calculate in such way that in
example wall lighted directly by a light source will light other objects
correctyl (acting a bit like light source with fade_power 2)?

3) It seem to be a problem to create a cololor that "only add" to backround.
In example it would be easy to create in OpenGL an sphere that will
be "adding white" - objects behind it will be lighten up by rgb 0.5 for
example. Using filter/transmit I seem to be able to MIX the colors, like
final_color = (background*X + sphere*(1-X))/2
so with proper filter the background color can be set to color of sphere,
but how to add color of sphere to background?
Media seem to be able to do that, but this is 3d, no way of doing it
(faslty) for 2d?



-- 

Post a reply to this message

Raf256 <spa### [at] raf256invalid> wrote:

> 1) blurred reflection - I see it is suggested (in FAQ) to use several normal
> patterns (or textures) and strong antialias.
> How about using the old method - because I think it might benefit in speed,
> by using adaptive number of samples (same as in camera focal blur). The AA
> of entire image is not as good probably

The old method?

There was an alternative method mentioned on here, once, although you'll
have to search for it if you're bothered. The idea was to do multiple
renders of the scene in which you want blurred reflection. A normal (based
on whatever pattern) is applied to each reflective surface you want
blurred. For each render, translate the normal by a different amount (most
easily by using the frame_number as input to a seed(), I imagine). Then
average together all the images in your paint application or using a
script.

You may find this method more visually satisfying during rendering since you
can usually see some sort of progress; in my experience, the single-pass
averaged blurred normal method has the same effect on povray as I imagine
superglued running shoes do on a marathon runner. But more effort is
required outside povray, and there are probably accuracy issues, which may
or may not be a problem depending on application. I have only used this
method once (and gave up totally on the approved method at that time), and
I didn't demand much of the effect, so you should really experiment
yourself.

> 2) radiosity and light fading - natural light behaves like with fade_power 2
> right? What about the radiosity? It is calculate in such way that in
> example wall lighted directly by a light source will light other objects
> correctyl (acting a bit like light source with fade_power 2)?

I imagine so; as I understand it samples are taken evenly (although
randomly) on the surface of a hemisphere. As you move a neighbouring
surface further away from the one doing the sampling, it will naturally be
hit by fewer samples, in fact the number of hits will decline by the
square, just like fade_power 2. So the effect is built-in, I think.

> 3) It seem to be a problem to create a cololor that "only add" to backround.

Slime came up with a way to do this ("additive transparency", is I think
what you're after) and posted it on this very news server. It worked very
well for ordinary transparent materials (no need for media), IIRC. See:

http://news.povray.org/4202c5f8%40news.povray.org

and

http://news.povray.org/4202c6ad%241%40news.povray.org


Tom

Post a reply to this message

Raf256 nous apporta ses lumieres en ce 28/05/2006 21:49:
> Hello,
> after ~year of beeing too bussy to use [seriously] PovRAY,
> I have few questions now when I try to go back to using it :)
> 
> 1) blurred reflection - I see it is suggested (in FAQ) to use several normal
> patterns (or textures) and strong antialias.
> How about using the old method - because I think it might benefit in speed,
> by using adaptive number of samples (same as in camera focal blur). The AA
> of entire image is not as good probably
You have 2 options here, both using averaged normal_map:
1 - some normal, scaled very small and randomly moved a small amount. Good with AA on.
2 - some normal, scaled large to very large randomly moved a large or very large
amount. Better 
result without AA.
You can combine the 2, using multiple pairs of normals, one small scale, one large
scale, but my 
testing showed somewhat unsastifactory results.
> 
> 2) radiosity and light fading - natural light behaves like with fade_power 2
> right? What about the radiosity? It is calculate in such way that in
> example wall lighted directly by a light source will light other objects
> correctyl (acting a bit like light source with fade_power 2)?
You don't have do do anything, it's automatic due to the way radiosity works. In fact,
a very long 
and narrow, bright object will have an effective fade_power of 1 as in real life and a
plane will 
have fade_power 0 (illumination independent to distance).
During my physics cources, I saw the demonstration and formulas showing that.


-- 
Alain
-------------------------------------------------
Everything can be filed under “miscellaneous”.

Post a reply to this message