RobF nous apporta ses lumieres en ce 18-02-2007 21:03:
> Hi,
> 
> Thought I'd post my latest WIP - an interior lighting study.
> 
> Overall, I'm pleased with the results, but I'm still having a few radiosity
> issues on which I could use some advice.
> 
> 1) The computer box seems too bright (there's no direct lighting on it, but
> it appears to glow) - setting to ambient 0 helped a bit.
> 2) Some splotchy shadows at the top of the bookshelves
also, an overly bright area to it's right. The wall behind the book shelves to 
the left is also to bright.
> 
> Any and all constructive criticisms are welcome.
> 
> Thanks!
> 
> Rob.
> http://fitzel.ca/dart/
> 
> p.s. here are the radiosity settings I used
>         pretrace_start 0.08
>         pretrace_end   0.004
reducing pretrece_end to 0.001 often allows you to use more relaxed setinge 
elsewhere.
>         count          400
you may need to increase the count, maybe around 800 to 1000.
>         nearest_count  8
going up to 15~20 forces you to average more values to get the final result. It 
may reduce speckling and splotchiness.
>         error_bound 0.7
this may be to low, try a value of 1 to 1.5, or even leave it at default value 
(1.8). It may reduce the splotchiness.
If error_bound is set to low, it can cause illumination of area that should be 
in the shadows, it also increase the render time.
>         recursion_limit 4   // need 4 if glass on window
>         low_error_factor 0.8
you may try reducing this to less than 0.5 (default)
>         gray_threshold 0.6
>         minimum_reuse 0.015
using a slightly lower value could help removing some lightign the should not be 
there. Be prudent, it can realy have a BIG effect on the rendering time.
>         brightness  2.5
>         adc_bailout 0.01/2
>         normal off
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
In a radiosity scene, it's good to add #default{finish{ambient 0}} and make sure 
that NO texture have ambient >0, unless the object is *suposed* to glow.
Try placing a black box just barely under the floor, or raising the computer a 
tiny bit, you may have "cracks" where your computer touch the floor that let 
rays reatch some bright area.

Personaly, I'd start by removing the glass pane during testing, then add it back 
in a two pass rendering.
Then, I would remark out the error_bound and low_error_factor as a first test step.
Next, try a higher count with a smaler pretrace_end like 0.002 or 0.001.
May try a low_error_factor of less than 0.5.

-- 
Alain
-------------------------------------------------
Caught Asleep At Your Work Desk

Just in case your boss catches you asleep at your desk, be ready to blurt out 
this excuse #1:  I was working smarter - not harder.

Post a reply to this message