I have made some tests with variable radiosity parameters. See
news://news.povray.org/3d6f88fa%40news.povray.org
for the discussion on this subject.
Varying count seems much more useful than varying error_bound. It's also
possible that changing error_bound depending on the position has some side
effects that are problematic. In any case the overhead produced by a
function based count value is much lower since the function only has to be
evaluated for every radiosity sample taken while an error_bound function
has to be called for every Compute_Ambient() call.
The following pictures show a fairly simple example. The first is
rendered with a fixed count of 60, the second with varying count. The
last shows the simple function used for defining the count: red means low
count (<20), yellow 20-50, blue >50.
The render times were:
126 seconds for fixed count
138 seconds for variable
about 13000 radiosity samples (and function VM calls), error_bound was
0.4, pretrace_start 1, pretrace_end 1/image_width, recursion_limit 2.
I think the difference is noticeable, notice especially the artefacts on
the outside of the cubicle. Anyway the parts with low count are looking
noticably worse too and the render time is longer. Testing with a real
scene and a better optimized function for the count value would be
important of course.
Christoph
--
POV-Ray tutorials, IsoWood include,
TransSkin and more: http://www.tu-bs.de/~y0013390/
Last updated 13 Aug. 2002 _____./\/^>_*_<^\/\.______
Post a reply to this message
Attachments:
Download 'var_rad_1.png' (70 KB)
Download 'var_rad_2.png' (74 KB)
Download 'var_rad_fn.png' (77 KB)
Preview of image 'var_rad_1.png'
Preview of image 'var_rad_2.png'
Preview of image 'var_rad_fn.png'
|