[argyllcms] Re: Newbie targen and colprof questions.
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Thu, 13 May 2010 17:00:15 +1000
Symphony Photography wrote:
1) colprof - When you finish generating an ICC profile it gives you a little
report such as:
"profile check complete, peak err = 17.211762, avg err = 0.458565, RMS =
0.657406"
Hi,
See the discussion at the bottom of the colprof page in the "Discussion"
section.
<http://www.argyllcms.com/doc/colprof.html>
of a read error is it possible to identify which line it was from, re-read
that line and update it in the ti3 file? I know the i1 Pro isn't the most...
errrmmm... consistent tool in the world so ideally I'll be averaging several
target scans together but if one of those is bogus I'd like to correct it
before it gets thrown into the pot.
If the peak error is really due to a misread, then maybe re-reading a particular
patch and strip may help. profcheck -v will list each patch and the
fit error. You will have to look in the .ti3 file to identify patch
locations though. You could try chartread -r to navigate about
and re-read a particular patch or strip. Note though that you are
likely to be wasting your time if the error is simply a reflection
of the fit error or inconsistency in the chart itself.
2) The targen -c argument makes reference to a "pre-conditioning" profile
but I can't find any reference as to the specifics of what a
This is discussed here <http://www.argyllcms.com/doc/Scenarios.html#PP2>.
3) I've been experimenting with various settings for target, and profile
generation but I'm having trouble with some of the mid-tones and shadows.
I've noticed this on a few different types of paper with both a Canon
IPF6100 and an IPF8300. Using the Bill Atkinson test page as an example if
I'm measuring a spot on the strawberries which reads 126,0,0 the profile
generated by Argyll will shift that to something significantly brighter. I
do not see this happening with using the 918 patch Bill Atkinson chart read
via the Measure Tool and converted using Eye One Match. This is clearly
visible on a Granger Chart as well. I don't think it's the instrument (i1
Pro) as I've used it with Eye One Match reading in the 918 patch Bill
Atkinson profile via the measure tool and density is fine (with the
exception of highlights, especially in the yellows). I'm pretty sure it's
just a matter of generating the right target (PEBKAC error.... Problem
Exists Between Keyboard And Chair) but I'm not sure where to go yet. Another
issue I'm having is with the specular highlights in the orange fall foliage
scene (in the first column of the Atkinson test image). The profiles I've
generated so far with Argyll seem to do away with all the specular
highlights, the result of which is a seemingly less detailed shot. Another
issue I'm seeing is certain areas of the profile that seem to have "waves"
of posterization although I'm pretty sure it's not posterization, it seems
to be in the profile. I know that when I was generating the ICC profile with
my initial tests with 462 path targets it was quite bad but introducing the
-qu argument in colproof went a long way to eliminate that.
You don't say what form of reproduction you are using. If it is perceptual
with gamut mapping then you may simply be seeing the results of gamut mapping.
The gamut mapping is influence by many things. For instance, check your viewing
condition settings, these can have an influence over the reproduction curve.
Highly
saturated colors get brightened up due to the Helmholtz-Kohlraush effect.
The V1.1.1 release does tend to preserve tonal distinctions at the cost
of contrast in the highlights, and it's something I'm working on
changing for the next release. Be wary of the Atkinson test images, as
they are rather strange if you look at their gamuts. Due to something
about their derivation or something like sharpening they have a lot of
energy in highly saturated colors with an L* value at 100. Such colors
are unreal, and therefore can be difficult to handle. An approach using
image specific gamut mapping combined with some heavy filtering
can help tame them. (something like tiffgamut -f80).
If you're not using perceptual, then I'm puzzled. Colorimetric reproduction
should be pretty close to instrument readings (taking into account chromatic
adaptation).
4) Just an observation - When using the i1 Pro to read targets, I've noticed
that I get a rather high dE from one scan to another when comparing
measurement files when using the Measure Tool. I was kind of surprised by
this as it's somewhat counter-intuitive to me. I figured that a slower scan
across the strips would result in more samples which would result in better
averaging and more consistent results but that seems to not be the case.
With Argyll however it is... if I scan as fast as I do with the measure tool
the reading will almost always fail. Argyll seems to like a REALLY slow scan
across the strips (fwiw I have the most recent version of hardware (just
bought it) so the sampling frequency should be rather quick).
I haven't noticed it needing particularly slow scans. The way the patches
were being sliced wasn't optimal in terms of accuracy though, and I've
changed this a little for the next release, so maybe this will
improve things.
5) Is it possible for Argyll to report DMax, and DMin either from doing a
few spot measurements or from an existing scanned target???
Not currently. It wouldn't be too hard to add it to spotread, although
there are so many flavours of density to deal with...
Graeme Gill.
Other related posts:
