|
|
Hi,
I am very interested in volumetric rendering, and since now POV have this
ability I try to play with it just a bit.
The fact is that I can only render volumetric fire or clouds that appears to
be "pointilized". I mea I can't get a continuos color inside the flame for
example. The fact is that if you take a photo of a flamne, the "interior"
appears continuos, so the question is,
there is some ways of taking away the pointilized effect?
Thanks a lot
--fabio
-----------------
------------------------------------------------------------
Fabio Pellacini
Ph.D. student - CS Dept. of Cornell U.
e-mail: fab### [at] cscornelledu
web-page: http://www.cs.cornell.edu/fabio
------------------------------------------------------------
Post a reply to this message
|
|
|
|
Hello
Fabio Pellacini wrote in message <36123626.0@news.povray.org>...
>Hi,
>I am very interested in volumetric rendering, and since now POV have this
>ability I try to play with it just a bit.
>The fact is that I can only render volumetric fire or clouds that appears
to
>be "pointilized". I mea I can't get a continuos color inside the flame for
>example. The fact is that if you take a photo of a flamne, the "interior"
>appears continuos, so the question is,
>there is some ways of taking away the pointilized effect?
>
>Thanks a lot
>
> --fabio
>
Try these, alone or combined :
+increase intervals (something like 15 or 20) ;
+increase samples : increase minimum to force samples; maximum to allow for
greater precision (I actually use samples 5, 25) ;
+increase confidence towards 1 (I use 0.9999)
+decrease variance (I use 1/255)
Of course there is a performance price to pay for each, so you'll need to
tune them.
I do _not_ know how it all works or what the parameters exactly mean, but
here's what I understood.
Intervals is the number of intervals taken on a line passing through the
media. Samples bound the number of samples taken for one interval (no less
then sample min ; no more then sample max). Confidence and variance are a
bit of statistical magic trick to decide how many sample to do, constrained
by samples bounds. I believe that variance is the (statistically predicted)
maximum acceptable difference between computed color and 'real' color (a
infinite number of sample). So I don't think any variance smaller then 1/255
is helping. Confidence is how much sure we are that we are whithin variance
compliance (also statistically predicted).
So, as I understand it, intervals works alone. For each interval, variance
and confidence are used to determine how many samples to use, this number
being bounded by samples min, max.
If you plan to animate your scene, you could use two settings. Animation is
less tolerant to the somewhat random process. So, settings that gets a
pleasant still image can still have an annoying flicker if animated.
I hope this helps,
Povingly,
Philippe
Post a reply to this message
|
|