-u+ option and noise

radiance-general
#1

Hi all,
I had already posed this question as a reply to the very interesting update by Mark Stock https://discourse.radiance-online.org/t/talk-a-decade-of-radiance-tricks/2712 but I thought it would have needed a post of its own.

In the slides by Mark Stock one reads that -u+ to could be used to remove noise and from the description of RPICT that this option would turn off Monte Carlo sampling. Would this be correct? Does this carry specific cons when used? When one should decide to use this option to reduce noise?

Thank you for your help!

#2

In fact, the opposite is true. Setting -u+ turns ON pure Monte Carlo sampling, in place of the low-discrepancy sequence normally used by rpict. The noise is still there, but with completely randomized characteristics. With the rpict default (-u-), you can sometimes detect patterns in the output, although it does converge slightly faster in that mode.

#3

Thank you Greg now it is clear!
So one should read:

Boolean switch to control uncorrelated random sampling. When “off”, a low-discrepancy sequence is used, which reduces variance but can result in a brushed appearance in specular highlights. When “on”, pure Monte Carlo sampling is used in all calculations.

with -u- (OFF, default) and -u+ (ON)

#4

If it helps, these boolean switches have synonyms. Any of -u-, -u0, -uN[o], -uF[alse], will turn MC sampling off. Any of -u+, -u1, -uY[es], -uT[rue] will turn MC sampling on. Use whichever is easiest/clearest for you. The -u option by itself “toggles” the setting to the opposite of whatever it was (normally the default, but multiple -u options switch back and forth).