Antoine Valentim found a bug in Mondrian filter. It should be
scale<1.2 * incrX, 1.2 * incrY, 0.01 * rand(lseed)>
instead of
scale<1.2 * incrX, 1.2 * incrX, 0.01 * rand(lseed)>
Also he recommends to use finish {ambient 1 diffuse 0}
It really works better.
Gena.
