[Radiance-general] Odd values when using gensky and ambient accuracy is zero...

[P George Lovell]

Hi Everyone,

Please excuse me if this question has come up before (or if it's simply 
a dumb question). I'm using mingw radiance with a matlab wrapper to make 
stereo stimuli for perceptual experiments. Up until now my illumination 
has been made by specifying a hemifield of small lights. I'm beginning 
to prepare an experiment which will involve the creation of a larger 
number of images, hence I'm trying to speed up my rendering. One option 
seems to be to use gensky to specify a more natural illumination field, 
this seems to halve my rendering times.

However, I only get satisfactory reproduction of the shadows when I use 
low values of ambient accuracy (not surprising). Shadows created with 
non zero ambient accuracy seem to be interpolated to a certain extent. 
Because my rendered images form a stereo pair, the differences between 
the shadows become very noticeable as irregularities in the surface 
depth, rather than as changes in brightness, this is a problem. I get 
much better rendering of the shadows when I  specify an 'aa' value of 
zero. But when I do this the reflectances seem to get messed up. The 
attached images include a series of monochrome tiles towards the bottom, 
these are linearly spaced in reflectance from 0.2 to 0.8. However for 
the gensky rendered image with aa set to 0 the darkest tile seems too 
dark. This jump in radiance disappears when I increase aa to 0.1, it 
also disappears when I change to multiple spherical lights and leave aa 
at zero.

Some example images:

http://i1235.photobucket.com/albums/ff431/GeorgeLovell/RadianceProblems/SkyType_sphere200Lights_aa_000_ar_0256.png

   - Things seem fine, though the rendering takes 780 seconds


http://i1235.photobucket.com/albums/ff431/GeorgeLovell/RadianceProblems/SkyType_gensky_aa_010_ar_0256.png

    - Rendering is quicker with gensky but with an aa value of .1 there 
are problems with the shadows.


http://i1235.photobucket.com/albums/ff431/GeorgeLovell/RadianceProblems/SkyType_gensky_aa_000_ar_0256.png

    - Here the darkest tile along the bottom of the image is too dark.

----
For reference: the leftmost and middle images form a cross-fuseable 
stereo pair, i.e. you cross your eyes when you look at them left eye 
looks at the middle image and the right eye looks at the leftmost image 
(this can give you a headache!). The rightmost image can be viewed using 
red/green glasses.
----

Does anyone have a clue why setting aa to zero seems to change the 
brightness of the darkest tile?

Thanks,

George



-- 
Department of Psychology, University of St Andrews

      RM 1.05 (The white room).
      Phone      (01334) 462090