GNU/Linux color management

Dear,

 

I have been trying (without success) to use color managed appllication under

GNU/Linux, mainly to be able to print on paper what I see on my computer.

 

All that with the use of cinepaint, gutenprint, lcms and lprog. No hardware tool

either since some people apparently can get accurate results without them, the

fact that the ones compatible with argyll cost a fortune is the other reason...

Also I am using adobe's free icc profiles.

 

Some people will probably advise to me to get a mac, I thought about it but I

have been using GNU/Linux for some time now and am quite familiar with it, color

management really is the only thing that gets in the way.

 

I have read several howtos on the subject but I haven't been able to get 1 print

that is close to what is displayed on the screen, generally the prints are brighter.

 

I would appreciate if some GNU/Linux users could/would like to share their

thoughts on the subject.

 

Thanks

Have you used a gamma tool like XGamma and a good chart like Norman Koren's to fix your gamma yet?

See <a

href="http://en.wikipedia.org/wiki/Linux_color_management">the

wikipedia article</a>, <a href="http://linux.vilars.com">this

site</a>, and maybe also <a

href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00HMf8">this

thread</a> on how a few things started.

Sure, from someone who's writing a RIP that uses LCMS...

 

First, you need that hardware color measurement device. No "by the eye" approach will even get you close, regardless of what "some people" tell you. Argyll supports both the old and new OPTIX (DTP92 and DTP94 respectively). A used DTP92 goes for under $75 on the bay. A brand new DTP94 is $129 at Midwest Photo Exchange (absolutely no connection to our own Midwest Photographic Workshops, tens of thousands of businesses in Michigan, Indiana, and Ohio are named "midwest" this or that).

 

http://www.mpex.com/ItemDetail.aspx?SKUID=10832&SKU=0834596000287&SN=0000&Q=False

 

Then, you need to control the lighting in your area. You really do need to "calibrate" a monitor by getting it to a decent black point and a white point that matches your room lighting (after it's filtered through your light fixtures and bounced off those ugly green walls that you swore to yourself you were going to paint last year. How do I know about your walls, you ask. I just do) and then you can profile the "monitor".

 

If you want to try the Gretag Eye One Display, that's in the same category as the Optix, the old version for maybe $75 used, the new one "Eye One Display Two" for $150 new. But you'll need to run the software under Windows, then load the profile into you Linux box, like in that article that Eugene pointed out. You could try booting the Linux box from a Windows XP CD set up with the BartPE environment.

 

http://www.nu2.nu/pebuilder/

 

Next thing you need is a printer profile. Color management is the systematic approach of editing the image file, viewing it through a good monitor profile, then printing it through a good printer profile. Now, making a profile on a Linux box isn't easy. Aside from needing a spectrophotometer ($500-$1500) you'll need profiling software that works on Linux (which probably means Windows software running under Wine). And you'll need practice. Your best bet might be to have a commercial profiling service do this for a reasonable price. There are places that will make a profile from $25 to $100, with $40 being a pretty common price.

 

I've never tried to get a monitor to print match on a UNIX system through your chain: cinepaint, gutenprint, lcms, and lprog (btw, what's "lprog"?) I have done it with Alastair Robinson's PhotoPrint.

 

http://www.blackfiveservices.co.uk/photoprint.shtml

Since color management by definition relies on accurate device profiles what you asking for is not quite color management. More specifically it's matching one monitor to one printer by eye. I don't know how to do it on any platform but I'm sure there's a way.

 

As far as color management goes, Eugene has mentioned the relevant links.

 

Essentially you'd use a profile created with a colorimeter on the same computer in Windows or OS X and use that profile in Linux (which usually requires installing one of the LUT loaders and configuring the colormanaged applications).

BTW, the cheapest colorimeter that is acceptable quality and is upgradable to very high end software is Spyder2Express.

Thanks, lot of usefull information there.

 

Steven Clark: Yesz I did without success but thanks!

 

Eugene Scherba: Yes I had a look there but the wikipedia description is helpfull to a certain extent only.

 

Joseph Wisniewski: This is good I didn't know there was a live WinXP cd. If you use photoprint then you use lcms (or little cms), it is a library on top of which programs run. Have a look there: http://en.wikipedia.org/wiki/LittleCMS

Oh and sorry it wasn't lprog but lprof... I am interested by how you seem to do things, you say that you print match under photoprint but isn't colour management supposed to be used under the photo editing software as well?

Are you sure about the Gretag-Macbeth? I mean are you sure that they need to be run under windows because they seem to be supported by argyll, but it might be a different model...

 

Serge Cashman: Hi the only way I know of to do it is to try and print with different settings from the gutenprint or whatever you use interface, you need patience, luck, ink and paper. And you might not even get there... I didn't!

Out of what Argyll lists as supported DTP-94 is the only one that still gets sold as part of the Optix bundle. The Eyeone bundles include a different colorimeter.

http://www.argyllcms.com/doc/ArgyllDoc.html

 

I think the three articles Eugene has linked to give all the info you need... Once you have the profiles for the monitor and the printer (and LUTs loaded if necessary) all you need is to configure each color managed application to use those profiles... Most of them use Little CMS so the option would look similar. Here's a screenshot from Scribus:

 

http://docs.scribus.net/content/en/screenshots/cmsmodify1.png

Damn right!

 

Only problem is that it's not possible to get that package where I live so I would need to order from the US. Add to that the fact that I don't really need a light meter and oyu get the idea...

 

I might have to wait for a second hand optix on ebay or go the wine route.

 

Also do you still use a LUT even though you have your profiles loaded?

It's just most calibration packages use LUTs to get to the targets you set. Then you have to load them after that.

 

In some packages you can profile a monitor without setting targets (for instance calibrate to Native/Native in Eyeone Display2). Then you don't need to load LUTs.

 

Or some lucky people (like Eugene with his Spectraview) have a setup that works via DDC, in which case you probably don't need LUTs either.

In your situation you can try just loading profiles from the monitor and the printer (or paper) manufacturers into the applications (you don't need LUT loaders for this). See if it gets you somewhere.

Just my 2 cents about using Wine (or Vmware) to calibrate your monitor : it didn't work for me (so please report back any success and how), because :

- USB support under wine is not good (never get the computer to talk to the device), but it was some time ago it might have changed since then

- Windows doesn't get access to the hardware (your monitor) and the software used for calibration can't calibrate : more specifically, in the case of the monaco products (Spyder2...) their software doesn't simply measure a set of color patch but change slightly bit after bit a profile that they reload after each measure... so for X it means changing the LUT. See the problem?

 

So you are back with having to perform all this *on the same machine* but with MS Windows and load the created profile under linux then.

Nicolas Vilars: Have you tried using Xen virtual machine?? You will need a quite recent to run windows under it though....

 

Joseph Wisniewskilso: some spectrophotmeter are supported by argyll, it seems to be a good idea to invest in one too since I might need several different profiles.

to Damien Mas: no I haven't try with xen although I have a recent laptop that support paravirtualization. I tried very hard to set up xen only to half fail: windows was working but a lot of pieces were missing to have anything decent as a laptop for work so I decided to wait six months for an other try.

 

That say, I'm not sure windows has access to the monitor/graphic card directly either. Perhaps if nvidia (my graphic ship on the laptop) does something that make it visible under the paravirtualized environment but since then...

Nicolas Vilars: funny that you say that because I was actually wondering if windows as a guest could have access to the graphic card.

 

A quick googling cleared this up.

 

On Xen Dom0 is your host machine and DomU your guest. If it was possible for your guest OS to access the graphic card, Dom0 would be the one to handle your it until you create a DomU. It is the transfer of the driver to DomU that seems to be delicate.

 

Have a look there: http://lists.xensource.com/archives/html/xen-users/2006-02/msg00449.html

I think accessing videocard during calibration is secondary to communicating with the colorimeter. If you don't have access to LUTs during calibration the software will just profile your display without them, which is not a big deal.

 

You can achieve most targets manually if you have a monitor with RGB controls. The only target you can't achieve without LUTs (or DDC) is Gamma. But unless you need to match the output of 2 monitors gamma really doesn't matter, you can just use native.

 

Match3 and Spyder2 software run in wine, but like Nicolas have mentioned they don't see their respective colorimeters...

to Serge: So was my thought, so was my thought... but that leaves two problems I've ran across.

 

- With vmware (USB is working), Spyder2Pro complained it hadn't access to a valid monitor (I suspect it was in fact the video card) -- something was missing and the software aborts the calibration before starting

 

- I tried doing the calibration with a second computer running MSWindows with the calibration software and capturing the color patch through a 24bits VNC session on my linux box. Although it produced a profile, this profile was "empty" => no visible color correction when applied to the monitor. I *guess* this is because the methodology Spyder2Pro use involves reloading a profile with slight color adaptation after each color patch is measured. Then, at least the spyder2 software requires access to LUT to do proper calibration.

 

The only thing that worked for me was to do the measure from a windows second hard disk on the same computer.

"...capturing the color patch through a 24bits VNC..."

 

Wow. That's quite impressive you went that far. Technically it can be a valid profile. You know, there usually are 3 stages in calibration: hardware adjustment, LUT adjustments and then profiling. Profiling being the important one and the other two more or less optional (although most software doesn't let you avoid the LUT adjustments). If a profile was created without LUTs then you don't need to load them later.

 

As long as the resulting monitor output was accurately measured and recorded the profile can be used. I would run some sort of validation though - like, compare measurments of patches sent via VNC to locally generated patches first and then run regular profile validation.

 

I think in your article you've mentioned that there's a possibility of loading separate LUTs to dual monitors. Do you know more details on that?

Actually for people in the US Joseph Wisniewski's link to DTP94 is quite a bargain. The problem is that it looks like Xrite is dropping those colorimeters in favor of EyeOne but there's a lot of other software vendors that support it. I've heard of the Midwest Photo Exchange, they cooperate with Strobist.com.