Random numbers / -DMC

Hi,
additionally to the general posting a few seconds ago: Reflections off specular surfaces showing patterns. - my two thoughts:
The visible patterns result from correlation between the sequence used in stratifying random numbers and the image width.
Either the stratification's sequence is enhanced (two orders of magn. longer ?) or substituted by something without a sequence.
Or (and more likely) the sequence is de-correlated with image width. A sequence in one line is ok, a sequence slightly one or two pixels off each line is (of course) visible.

-Peter

···

--
pab-opto, Freiburg, Germany, www.pab-opto.de

The correlation with pixel position is intentional. Low-discrepancy
sequences necessarily have some correlation, and it is up to the
programmer to decide how this manifests itself. If you don't choose
something, you are likely to be surprised and disappointed with the
results, as I was initially. I selected this pattern based on its
"brushed" appearance. There is a function (pixnumber?) in src/rt/rpict.c
that determines this, and you are welcome to play around with it.

-Greg

Quoting Peter Apian-Bennewitz <[email protected]>:

···

Hi,
additionally to the general posting a few seconds ago: Reflections off
specular surfaces showing patterns. - my two thoughts:
The visible patterns result from correlation between the sequence used
in stratifying random numbers and the image width.
Either the stratification's sequence is enhanced (two orders of magn.
longer ?) or substituted by something without a sequence.
Or (and more likely) the sequence is de-correlated with image width. A
sequence in one line is ok, a sequence slightly one or two pixels off
each line is (of course) visible.

-Peter

[email protected] wrote:

The correlation with pixel position is intentional. Low-discrepancy
sequences necessarily have some correlation, and it is up to the
programmer to decide how this manifests itself. If you don't choose
something, you are likely to be surprised and disappointed with the
results, as I was initially. I selected this pattern based on its
"brushed" appearance. There is a function (pixnumber?) in src/rt/rpict.c
that determines this, and you are welcome to play around with it.

So what we really need would be some 2-dimensional random number
generator producing a less conspicuous screening pattern?
Assuming that such a thing even exists, of course...

-schorsch

···

--
Georg Mischler -- simulations developer -- schorsch at schorsch com
+schorsch.com+ -- lighting design tools -- http://www.schorsch.com/