[argyllcms] Re: profile <1 million dollar question!>
- From: "Alastair M. Robinson" <profiling@xxxxxxxxxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 29 Feb 2008 20:27:03 +0000
Hi
Roger Breton wrote:
A quick comparison turns out a very good fitting of the device data by
Argyll. Very good. For example, for the 100C patch, the EyeOnePro measured
51.499 -34.625 -57.789 whereas the TIFF file opened up in Photoshop shows
52 -35 -57.
Impressive.
But then I turn to the comments by profile and I think to myself, maybe
there are some profile generation statistics I missed that would tell me how
good was the fitting of the colorimetric data by the profile? And the only
thing I can think of is the very last line:
profile check complete, peak err = 4.363131, avg err = 0.961138
Yup, that's precisely what it does. It's a measure of how well the data
in the profile LUTs fits with the measured data. If you want to find
those numbers for an existing profile (and its .ti3 file), by the way,
you can use the profcheck command.
can I reasonably expect those two statistics to go down?
I don't think you can count on it. If my memory serves, from when I
asked a similar question, it's actually easier to achieve a good fit to
a sparse set of data points; when the number of patches is low, it's
possible for there to be "wrinkles" in the device response which are
essentially glossed over by the profile. It's possible that using a
larger number of patches might reduce the peak error, but it's also
possible the average error might go up. It depends on the device.
Or is that a sign of the device ill-behavior?
Well, it's that too. Device ill-behaviour is the most likely reason for
a bad fit!
This is from an Epson printer, by the way. Linearized by a CMYK RIP.
Well, those numbers look reasonably typical for an inkjet - in the
profiles I generate (which are usually 400 or 800 patch RGB) I tend to
see between about 2.5 and 4.5 for max, and between 0.5 and 1.5 for
average, depending on media.
All the best,
--
Alastair M. Robinson
Other related posts:
