Re: Profile causes posterisation
Re: Profile causes posterisation
- Subject: Re: Profile causes posterisation
- From: Chris Murphy <email@hidden>
- Date: Mon, 30 May 2005 15:44:09 -0600
On May 30, 2005, at 2:54 PM, Stone Quay Studio wrote:
To further complicate things, before coming to me, this client had a
profile
made by one of the big London colour management specialists paying
£135 for
a profile (same paper, ink, printer) made with ProfileMaker and a
spectrolino. That profile does the exact same thing in the exact same
places
(he brought it back to that specialist who spent an hour with him and
told
him to get his camera profiled). However, a profile this client had
made for
him a few years ago doesn't cause the posterisation effect nor does the
epson canned profile though both leave a lot to be desired in other
areas.
What driver settings were used to build all three profiles? If you and
the specialist have built the profile with "No Color Adjustment" and
the "few years ago" profile was built with Photorealistic, this could
easily explain the difference in behavior. How many patches were read
for the building of all three profiles? How many samples were taken and
averaged? Was the data verified with at least two measurements of the
target (either the same target or a copy) to ensure the data is valid?
What's the size of each of the three profiles? I'm not sure if the
latest versions of i1 Match are building Large profiles by default or
not, earlier versions were not building large profiles. ProfileMaker
can and should be set to make a Large profile.
Part of the point here is that you may have no idea what the answer is
to all of these questions, in particular with respect to a three year
old profile.
However, it's possible with very non-linear output devices, in
particular the way the No Color Adjustment tables in Epson drivers
behave, that could result in posterization as you describe. I've had
this happen with an unrelated devices quite a few years ago where a
small number of patches (81 I believe), would produce a very smooth
profile but he shadow and highlight detail were getting hit, and if you
wanted to do a color match of some key colors there would be an obvious
discrepancy. Considering only 81 patches were used, it worked well and
that's what the client ended up using. Using about 400 patches, the
posterization set it and it was also obvious but in this case also
unacceptable.
My suggestions:
1. Track down another 2100 using the same media and make a profile for
it. This will eliminate your friend's 2100 from being the source of the
problem - it may in fact be behaving wonky.
2. Use synthetic targets such as the Granger Rainbow, and Don
Hutcheson's RGBXPLORER8.tif, and also the LAB ramp. Print as you would
any other image, using Photoshop print with preview to convert using
the suspected problematic profile, and the proper driver settings. If
you're getting a whacked print with one of this synthetic targets, it's
the printer profile or it's a seriously misbehaving printer. These
targets can be found here:
http://www.colorremedies.com/realworldcolor/downloads.html
http://www.hutchcolor.com/Images_and_targets.html
3. Produce a new profile using PhotoRealistic instead of No Color
Adjustment, since this setting is much more well behaved. You will give
up some gamut however.
4. Produce a new profile using a larger number of patches. As in 2500
or so. It's possible the number of patches you're using now is causing
a bad profile to be built, that's causing profile induced
posterization. Using something like 2500 patches to brute force unwind
whatever non-linearity is causing this problem may be necessary. The
profiling process sees this non-linearity WHERE IT HAS MEASUREMENTS and
tries to compensate for it, making assumptions about the behavior
between the patches which end up being wrong, and that causes the
posterization. Measuring substantially more patches can sometimes solve
this problem. (The easier solution is to measure substantially fewer
patches - but then that makes for a far less accurate profile, but it
is a valid option.)
5. As for the expert saying the solution is to get the camera profiled,
that sounds like bunk, but test #1 and #2 should isolate whether it is
in fact the profile causing the problem. If so, hopefully the expert
will help your client out, or refund the profiling fee as it sounds
like the current profile he's made (and yours for that matter) are of
no use to your client.
Chris Murphy
Color Remedies (TM)
www.colorremedies.com/realworldcolor
---------------------------------------------------------
Co-author "Real World Color Management, 2nd Ed"
Published by PeachPit Press (ISBN 0-321-26722-2)
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden