[argyllcms] Re: Very poor results with 1000+ patch target.
- From: Kamil Tresnak <kamil.tresnak@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sun, 12 Oct 2014 22:45:03 +0200
Hello Michael,
what Brad suggested is good way, it is discussed in Argyll doc in more
detail. Also, you can try -s parameter, which give you some single channel
steps, including yellow. Last, but not least, you can try something
uncommon, but as you are IT guy, this will be pretty easy for you - you can
add some patches to TI1 file manually. Of course, you have to edit all
others parameters, like number of data sets etc, but i think, you can do it.
Acc to -p: this parameter is good to tell to targen if you want more
patches in shadows or in lights.
Good luck!
Kamil
On Sun, Oct 12, 2014 at 9:42 AM, Michael Gallagher <gallaghermikey@xxxxxxxxx
> wrote:
> Thanks for the continued help everyone! Kamil, one of the colors that
> seemed to be lacking after creating this profile were the yellows. Could
> you elaborate on what I would need to do in order to optimize the profile
> for yellows, specifically? I'm a bit lost on how to use this power
> parameter you referred to that allows me to improve the colors that are
> important to me. I did read the documentation on targen but the -p flag is
> still a bit confusing to me.
>
> Thanks again.
>
> Mike
>
> —
> Sent from Mailbox <https://www.dropbox.com/mailbox>
>
>
> On Wed, Oct 8, 2014 at 4:04 AM, Kamil Tresnak <kamil.tresnak@xxxxxxxxx>
> wrote:
>
>> Hello Michael,
>> congrats!
>> So problem was in measuring procedure?
>> As a next step, if you wish experimenting a make some progress, i would
>> recommend chart with little bit more patches, you probably do not need
>> thousands of patches, but somewhere around 1500 can give you more satisfied
>> results (see Argyll doc, you can experiment with power parameter and cover
>> colors which are important for you).
>> Regards,
>>
>> Kamil
>>
>>
>> On Wed, Oct 8, 2014 at 6:18 AM, Michael Gallagher <
>> gallaghermikey@xxxxxxxxx> wrote:
>>
>>> Success! I made a 524 patch sheet, and using the ruler and slit I was
>>> able to get this on running colprof:
>>>
>>> Peak err = 1.633284, avg err = 0.377664, RMS = 0.441556
>>>
>>> Here's the IT3 for anyone interested:
>>> https://www.dropbox.com/s/yjdn3a8lxyls2h8/Arches_524_i1.ti3?dl=0
>>>
>>> I'm assuming this is pretty good, right? Either way, the resulting test
>>> print was, by far, the best that we've seen. My wife and I are very happy
>>> with it. Thank you all for the help. The next step will be to figure out
>>> the scanner's accuracy, since that does seem to be off slightly.
>>>
>>> Kamil and Alan: It turns out the ColorSync app DOES correctly print out
>>> targets (of course, when "Print as color target" is selected). To test this
>>> I cropped a strip of patches from a target, and printed two of them side by
>>> side. One with ColorSync and the other with Adobe CPU. When comparing the
>>> two rows, there is no difference between them at all.
>>>
>>> Matthew: we're in northern Utah. The processes you mentioned sound very
>>> interesting, and I would love to see that PDF. My understanding of color
>>> matching is quite limited (my wife is the artist, but I'm a CS student and
>>> I'm comfortable on a command line so figuring out ArgyllCMS became my
>>> thing), but I am interested in seeing how others do manage to do stuff like
>>> this. Do you coat your papers, or use inkjet watercolor paper, or anything
>>> of the sort? One thing I'm noticing is that the thin lines don't seem to be
>>> quite as sharp when printed. It's not that bad though. I'm assuming this is
>>> just because I'm printing on a paper not meant for printing.
>>>
>>> Graeme: Thank you very much for your suggestions. A couple questions: is
>>> there much to be gained from optimizing the profile I have? e.g. running
>>> targen -c with my new ICC profile. Also, after running colprof -v, the ICC
>>> file shrinks in size. It went from 1.2 MB to 279 KB. Is that profile still
>>> usable?
>>>
>>> Thanks again,
>>>
>>> Mike
>>>
>>> On Mon, Oct 6, 2014 at 1:11 AM, Kamil Tresnak <kamil.tresnak@xxxxxxxxx>
>>> wrote:
>>>
>>>> Hello Michael,
>>>>
>>>> maybe, the problem is in the printed charts, maybe there is a
>>>> difference between printed chart and ColorSync PDF export. Your
>>>> printed chart - seems like OK? Maybe you can find some colors which
>>>> are exactly same (CMYK source valueas) on both charts (maybe full
>>>> c/m/y) and make spot color measuring to ensure that you dont have
>>>> problem with printing output.
>>>>
>>>> Regards,
>>>>
>>>> Kamil Tresnak
>>>>
>>>>
>>>> On Sun, Oct 5, 2014 at 4:01 AM, Michael Gallagher
>>>> <gallaghermikey@xxxxxxxxx> wrote:
>>>> > Hmm... here's something I just noticed. In ColorSync, when I select
>>>> "Print
>>>> > as color target", the colors in the little preview window end up
>>>> changing
>>>> > drastically. Is this normal? I saved the preview as a PDF on the
>>>> bottom of
>>>> > the Print window to show you guys. Here they are:
>>>> >
>>>> > Original Argyll generated TIF target:
>>>> >
>>>> https://www.dropbox.com/s/kfp78d7oytepynh/1050_Arches%20Original.tif?dl=0
>>>> >
>>>> > ColorSync Target saved as PDF:
>>>> >
>>>> https://www.dropbox.com/s/b6pvvsgniw9misz/Arches%20ColorSync%20Target.pdf?dl=0
>>>> >
>>>> > I'm going to try using Adobe's Color Print Utility to see if the same
>>>> thing
>>>> > happens during the print.
>>>> >
>>>> > On Sat, Oct 4, 2014 at 7:39 PM, Ben Goren <ben@xxxxxxxxxxxxxxxx>
>>>> wrote:
>>>> >>
>>>> >> On Oct 4, 2014, at 6:23 PM, Michael Gallagher <
>>>> gallaghermikey@xxxxxxxxx>
>>>> >> wrote:
>>>> >>
>>>> >> > Forgot to ask again: After I make this new target, is there any
>>>> way to
>>>> >> > generate targets/patches from a JPEG to optimize the profile?
>>>> >>
>>>> >> That sort of thing is much more useful for input (camera or scanner)
>>>> >> profiles. In your case, what you'll want to do is use the "basic"
>>>> profile
>>>> >> you're about to create to "pre-condition" the patch generation
>>>> algorithm in
>>>> >> targen; see the "-c" option.
>>>> >>
>>>> >> b&
>>>> >
>>>> >
>>>>
>>>>
>>>
>>
>
Other related posts:
