[argyllcms] Re: Printer grey-scaling profiling inaccuracy (using scanner)
- From: Milton Taylor <milton.taylor@xxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 24 Mar 2006 21:51:20 +1100
Hi Gerhard,
what do you mean with "gray values"? Which "values" do you call
"gray"? Do you mean printer specific DeviceRGB values, where R=G=B?
Notice, that particularly for device specific color spaces it is not
necessarily granted, that R=G=B generally applies for neutral colors,
though in your case, this likely happens to apply (at least with
original consumables), since the manufactorer has likely
implemented/calibrated the driver in a way such that it
accepts/expects sRGB input.
Yes, to reiterate, I did mean device specific values where r=g=b, and
yes I understand that this doesn't guarantee grey, but hopefully will be
close if the manufacturer did their job well!
as inputs to the printer it does indeed generate visually neutral
greys on the paper. This is not really surprising as the printer
should only be using black ink at that point,
I would not necessarily expect that. Very likely, the driver does not
print neutral colors with black ink only, particularly not lighter
gray tones.
Well you may be right about that. ...I was really looking more closely
at the midtones, and under the loupe I could not see coloured dots so i
assumed that it was using black ink only.
Looks like you are experiencing metamerism, i.e. you're experiencing
the limitations of using a scanner as poor man's colorimeter.
Yes I think so! I put the IT8 and the printouts on the flatbed
together, and even though they both look very similar and quite neutral
grey to my eye, clearly the scanner is seeing them differently.
Yes, even if the scanner "sees" the same RGB numbers, when it "looks"
at the printed gray ramp and at the IT8 gray ramp, the eye can still
see these colors differently (and vice versa). The scanner "sees"
colors in a different way, than we do.
So it would seem! I would appear to have a case of the 'vice versa',
i.e. same appearance yielding different RGB values.
Coming back to an earlier question of mine then, I guess that really
does mean that a flatbed scanner profile is only really accurate if it's
used to scan from the same type of paper and ink as the target was made
from. I don't know why this is all that surprising since it's certainly
true of film scanners...e.g. you can't scan Velvia film accurately
unless you made a profile from a Velvia target.
i.e. I assume that if I had a (measured) target that was actually made
on my printer/paper/ink combination, that it would actually produce
better results, not far off what a spectrophotometer would produce...
In the mean time, I have two possible solutions to this, both of which
are less than scientific, and based on visual guess work. (i.e that I
trust my eye to pick a neutral grey) However it just might work:
a. If the scanner is seeing all colors consistently too red say, I could
actually post-adjust the image of the scanned pizza chart in Photoshop,
so that the values for the grey scales end up neutral again, by winding
back the red channel until the greys are indeed neutral [in device
space] in the resulting file. In effect I would be compensating for the
way the scanner seeks the printer ink/paper. The modified tif's would be
then fed into 'scanin', and then to profile.exe. This is only going to
work if I can be reasonable sure about what the scanner is actually seeing.
b. I did use the -g option when I created the print chart. If I remove
the grey wedges from the file, the profile will presumably not include
LUT adjustments for these grey scale values, and so the profile will
more or less leave the greys alone when they go through from PCS to
printer device space. But that assumes that this metamerism effect in
the scanner is affecting the neutral greys more so than the saturated
colors. I have a suspicion this might be the case, as the visual match
in the colors is actually not too bad. I think this will work, because
in an earlier test I left out the grey wedges off the printed target,
and did manage to get some neutral greys out of the printer. At that
point I really didn't quite have a handle on what was going on, so it
did not click then.
Oh well, I have reached the limits of the technology currently at my
disposal. :-)
Cheers,
Milt
Other related posts:
