 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
http://195.221.122.126/samples/erreur.html
M
Post a reply to this message
|
|
| |
| |
|
|
From: George Pantazopoulos
Subject: Re: Rad_data vs Halton , a quantitative comparison
Date: 7 Jan 2003 02:18:54
Message: <3e1a7f5e$1@news.povray.org>
|
|
|
| |
| |
|
|
Very interesting Mael. Is your Halton sequence a precomputed, drop-in
replacement for rad_data[1600] ?
George
"Mael" <mae### [at] hotmail com> wrote in message
news:3e1a7ec5$1@news.povray.org...
> http://195.221.122.126/samples/erreur.html
>
> M
>
>
Post a reply to this message
|
|
| |
| |
|
|
From: Christoph Hormann
Subject: Re: Rad_data vs Halton , a quantitative comparison
Date: 7 Jan 2003 04:09:09
Message: <3E1A9934.17AB19A0@gmx.de>
|
|
|
| |
| |
|
|
Mael wrote:
>
> http://195.221.122.126/samples/erreur.html
>
Intersting, your halton sequences always seemed too random for good
results to me. Does the difference also show up in the actual renders?
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 31 Dec. 2002 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
From: Mael
Subject: Re: Rad_data vs Halton , a quantitative comparison
Date: 7 Jan 2003 04:22:31
Message: <3e1a9c57@news.povray.org>
|
|
|
| |
| |
|
|
> Intersting, your halton sequences always seemed too random for good
> results to me.
May be some sequences can do better, if you send me your set of points I can
try to add the curve I get with it
However monte carlo methods does work with random sequences, and quasi monte
carlo with pseudo random sequences (and in pseudo random there is still random
:)
> Does the difference also show up in the actual renders?
I've not tried visual tests. Note that the 0.2 difference between 1600
rad_data and 1600 halton if equally dispatched on the 1053 locations only
represents a small amount of illuminance ..
I think I should also have traced the curve for the max error
M
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> http://195.221.122.126/samples/erreur.html
Updated with curves for a 300 points set by Christoph Hormann, and also added
curves for the max error
The scene rendered for 50 samples for the three distributions
http://195.221.122.126/samples/erreur_images.html
this is low quality radiosity :
1 pass (about 1000 points where illuminance is gathered), high error_bound
(but the aim was to evaluate the quality of the gathering part, not the whole
radiosity algorithm :)
M
Post a reply to this message
|
|
| |
| |
|
|
From: Thorsten Froehlich
Subject: Re: Rad_data vs Halton , a quantitative comparison
Date: 7 Jan 2003 09:19:35
Message: <3e1ae1f7$1@news.povray.org>
|
|
|
| |
| |
|
|
In article <3e1acbf1$1@news.povray.org> , "Mael" <mae### [at] hotmailcom>
wrote:
>> http://195.221.122.126/samples/erreur.html
>
> Updated with curves for a 300 points set by Christoph Hormann, and also added
> curves for the max error
>
> The scene rendered for 50 samples for the three distributions
> http://195.221.122.126/samples/erreur_images.html
> this is low quality radiosity :
> 1 pass (about 1000 points where illuminance is gathered), high error_bound
> (but the aim was to evaluate the quality of the gathering part, not the whole
> radiosity algorithm :)
Pardon my ignorance, but I see no real difference in any of the
pictures...???
Thorsten
____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde
Visit POV-Ray on the web: http://mac.povray.org
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
A little precision about those graphs :
The curves don't give any indication about the visual aspect of the images.
They only measure a distance to an 'exact' solution.
Particularly for low count the pov image can be smooth and look good (but be
far from the exact solution) and the halton image have artifacts (but be
statistically closer to the exact solution)
M
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
