> This effect is cause by the clipping of prediction values (stored
> illuminance values plus gradients) in ra_average_near() (line 461 ff. in
> radiosit.cpp).  I am not completely sure why this clipping is used but i
> suppose it was meant to reduce artefacts when there was no
> Maximum_Sample_Brightness yet.  Using max_sample you can achieve about the
> same effect.

Nope :)

You simply do not understand gradients in the Ward's method....
It does not limit the interpolated values.. it limits overestimation by
gradient method and
has no connection to the Kari's sample.

I will show you in a few days all clipping places in the radiosit.cpp and
more :)

Post a reply to this message