[argyllcms] Purplish blues with Perceptual in Argyll 1.1.0_RC1

  • From: Elena [service address] <1007140@xxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 26 Nov 2009 17:38:07 +0100

Dear Graeme, dear list,
I'm a new user just approaching Argyll. I'm from Italy, so please allow
my perhaps poor English.

Just to introduce me, I'm currently using an Epson 4800 driven low-level by 
software written by myself. Yes I'm quite expert with these stuff. My software
provides weaving, screening, linearization and all the rest. Ok that's not the
point of course, it was just to understand that I can really drive my printer
with full freedom and total control, ways over a good RIP could go.

Until now I was using ProfileMakerPro 5 and EyeOnePro to profile my workflow 
good results, and usually Photoshop for profile conversion and preview. I 
decided to give a try to Argyll, to see if even better results and flexibility 
be achieved. Yes I know the learning curve is steep, and it will take me time
and efforts to master all the relevant parameters.

The first experiment I made with Argyll is just to see how good it generates
CMYK output profiles (please note that I don't understand much of device link
and abstract profiles, I never used them...). So I started converting a
782 patches testchart reading I already had at hand, made with MeasureTool,
into Argyll format with logo2cgats utility and passed it to colprof to
create a profile with the following command line

 colprof -v -qm -kp 0 .3 1 1 .5 -l300 -L100 -al -S sRGB.icm Elena782

just to start. Note that the ink limit and K generation parameters quite match
the ones I usually use. I also understood from Argyll documentation that I have 
specify a source profile gamut in order to generate perceptual intent tables, 
so I
specified SRGB since the pictures intended for the tests are SRGB. I generated 
profile and suddently printed some test pictures to evaluate it, both with Perc
and RC intents. Here's what emerged:

1) With Perceptual, the RGB pure Blue comes out really purplish rather than
   blue, a very annoying thing. With RelCol, it comes out plain blue, i.e
   what it should be (sorry if I can't attach a scan, my scanner should
   be profiled better still, currently it isn't able to show very much
   this purple cast in blues).

   Curious, I made the same test with the previous release of Argyll binary
   (1.0.4), very same command, to find out that now the blue in Perceptual
   came back blue, no more purplish! Is it a bug of the current RelCand. ?
   Or something has been changed and I missed it ? Where am I wrong ?

2) What is good, it seems that Argyll can actually rebuild the black generation
   much better than PM does, even if the test char was optimized for a different
   generation. In my example, the 782 patches test chart created with Measure 
   was optimized for GCR 4, K start at 0. If you notice the command line above,
   I tried to delay K start to 30%, and the grey scale coming out from Argyll
   profile is just one word: perfect. Note that if I try to do the same with
   ProfileMaker, i.e forcing a K start 30% from a chart optimized for K start 0%
   or vice versa, the resulting grey scale has slight color bands.

3) A pity, Argyll profile still has "bumps",or "waves", expecially noticeable 
   printing out fully saturated RGB test fades, HUE circles and such things. If 
   don't understand what I'm meaning, it's the same problem discussed in
   a past topic

   Yes, they are those waves whose frequency increases if you increase the
   LUT resolution. I noticed no major improvement in this problem from
   1.0.4 to 1.1.0. Btw, ProfileMaker too has this defect, perhaps a bit
   less noticeable, but it's there. I wouldn't stress you too much on this
   problem because I figure out that using a much bigger testchart (perhaps
   thousands samples rather than 700-800) the situation could improve. At
   least, I hope so...
   Oh, also I actually was unable to go beyond -qm because with -qh my PC is
   still calculating the profile after several hours (don't know if that's
   the expected behaviour... it's a Win XPPro machine with P4 dual core 3Ghz, 
   really crap...)

Thanks for your attention and forgive my long email!

Other related posts: