On Wed, 22 Feb 2006, Gerhard Fuernkranz wrote:
Cory Papenfuss wrote:
Yes... way better, and without the numerical weirdness of the local
convergence. I've now only got a couple of patches that have perceivable
dE in iccexamin, and they're probably due to a glare on my test shot. No
more low-luminance progressive errors. I'm a little curious as to what the
two new "knobs" to tweak the spline fit do. I've so far left them at
default.
The sliders contol the trade-off between fitting error and smoothness. Moving
them up, will increase smoothness, but also the fitting error.
The two sliders correspond to the -r<avgdev> and -rs<smooth> options of the
Argyll CMS' profile command (see http://argyllcms.com/doc/profile.html), with
the exception that lprof's Avgdev = <avgdev> / 100.
Both sliders map to a single smoothness/error trade-off metric internally, so
they are basically redundant, but have a different scale. I think that one of
the sliders will be obsoleted eventually.
Thanks for the info. The only parameter that I played with in
argyll that appeared to affect the output was was the -a flag.
People usually tend to desire highly accurate profiles, and mostly measure
the accuracy simply by comparing the profile to the same data set that was
used to create the profile (that's alo what happens if you open the profile
checker window after creating the profile). This is IMO mathematically not
well-founded. First, even a perfect fit to the measurements used to create
the profile does not imply that the profile would fit as well for a different
data set. Second, measurements are never error-free, but always (sometimes
more, sometimes less) noisy. If you create a profile from measurements which
have a random error (noise) of say 2dE average, and if the profile fits this
data set with say 0.3 dE average, then it is IMO very likely, that the
profile is not really as accurate as it seems; it rather fits the noise in
the data, instead of describing the avarage behaviour of the device more
accurately. Given such noisy measurements, a smoother profile with a larger
fitting error may actually describe the average behaviour of the device more
accurately, than an supposedly accurate profile, with a low fitting error.
Regards,
Gerhard
Interesting thought. Especially since the gamut of the fit color
space is much wider than what is measureable from the target. It's
impressive how much larger the space is than the target, so if you've got
a "super-accurate" fit to the data, it could well be goofy outside.
Just a thought about the noise. I know that the whole Bayer thing
throws a wrench in the works, but maybe some median filtering rather than
simple averaging? I suspect that it's mostly evident at the low-levels
where SNR is lower. That's where the pixels that are "hot" matter most.
-Cory
--
*************************************************************************
* Cory Papenfuss *
* Electrical Engineering candidate Ph.D. graduate student *
* Virginia Polytechnic Institute and State University *
*************************************************************************
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642
_______________________________________________
Lcms-user mailing list
Lcms-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lcms-user