I have done extensive and I believe that Photoshop displays wrong colors when running on a PC with a calibrated monitor (see the full at: http://antoine.garric.free.fr/photoshop_colors/). I wonder if others have seen the same issue and how they solved this issue.
My explanation so far is that Photoshop knows that a profile is installed and modifies the colors to be displayed. But then Photoshop passes the image to the underlying operating system for display. Windows applies the same correction, resulting in wrong colors (the image is modified twice).
Any help on this subject is welcome.
I'm not going to pretend that I understand all the technicalities here, but if you want a screenshot from Photoshop to show the colors exactly as you see them on screen, you need to first ASSIGN your monitor profile, then convert that to whatever space you're using (like sRGB for web). I'm not sure that's necessary from a non-color-managed app, so that could account for the difference.
Windows does no color correction for display. Only Photoshop (or another color managed application) corrects the colors for the display.
If Photoshop is not showing you the correct colors, then the only possibilities are that the display profile is incorrect, or the display is broken. (usually it's the profile)
D Fosse says: if you want a screenshot from Photoshop to show the colors exactly as you see them on screen, you need to first ASSIGN your monitor profile.
Thanks for reading my question. I would like to explain why I do a screenshot:
My global problem is to explain why a sRGB file is looking different between Photoshop and Microsoft tools. With my eyes I can see differences but it is sometime difficult to be sure. So to make sure if images are identical or different I made a screenshot and looked at the values in the file. Do you think this is valid thing to do?
Regards,
Antoine
Chris Cox said: Windows does no color correction for display. Only Photoshop (or another color managed application) corrects the colors for the display. If Photoshop is not showing you the correct colors, then the only possibilities are that the display profile is incorrect, or the display is broken. (usually it's the profile)
Hello Chris, thanks for looking at my question. May be I am wrong but I believe that windows DOES color correction. The day I calibrated my monitor with Eye One, this software created a profile, this profile was installed in windows and immediatly the colors changed on my screen it used to be blueish, it is now more brownish.
So this is a color correction from windows. Isn't it?
This color correction should make my sRGB files viewed with microsoft tools look closer to real colors. If it is not the case what is the purpose of this calibration?
regards,
Antoine
Chris thanks again for your time looking at my question. See my remarks below:
Chris Cox said : Last time I checked, Windows FAX viewer was not color managed.
I am not sure what you exaclty mean by color manage but I observed that my 2 gretagMacbeth files in sRGB and Adobe98 profiles are displayed the same. To me it means that with no doubts the Fax Viewer utility takes the profiles into accounts and makes the appropriate corrections. If you had to this the fact that my monitor was calibrated and profiled and that windows corrected the colors acordingly (see my previous post) then I don' see what this tool is missing to display the right colors???
Chris Cox said : And your analysis assumes that your display profile is correct.
I suspect you just have a bad EyeOne, or a bad version of the EyeOne software.
Well I use the one eye from my club and others have been happy with it. But even though it might be wrong this is not the question.
My real question is why are the colors different between Photoshop and Microsoft tools? Even with a bad profile, both Photoshop and Windows have the same calibration info so then should end up display the same wrong color. Why isn't it the case?
Regards,
Antoine
agarric wrote:
My global problem is to explain why a sRGB file is looking different between Photoshop and Microsoft tools.
Regards,
Antoine
Photoshop is a fully colour managed application. It understands and respects any ICC profile tagged in your image and also uses the monitor profile to display the image. Few other windows applications, including those from Microsoft, have complete colour management.
I downloaded your file and it appears far different in Photoshop on my machine than your posted version. You almost certainly have a bad monitor profile.
Please delete your current monitor profile and recalibratre. All fully colour managed applications will display your image identically; every other application will show differences. And if you use a high gamut monitor things get even trickier if using non-colour managed applications.
Wolf Eilers said:
Also, for an application to be fully colour managed it must use the monitor profile.
It is not sufficient to just have the OS be aware of the profile.
Wolf thanks to look at my posts and giving me advises.
May be you can explain one thing to me:
Lets forget Photoshop and other sofisticated applications and also lets forget Adobe 1998 for a moment. Let just talk about pain, fax viewer, etc... diplaying sRGB pictures.
What is the purpose of doing calibration and installing a profile within Windows if this is not to calibrate properly you Monitor? I thought that by doing that I would have all standard applications displaying properly sRGB file.
Your post seems to say that this is not the case, that this is not enough to get the right colors. Ok but then why do we calibrate if this is to get colors that are still wrong at the end?
When installing this profile in Windows (for example with eye one software), on my PC the colors change dramticaly. The display used to be a bit blueish and now become more reddish.
Isn't it that the calibration is applied?
Why do you say that this is not enough? Why would the Eye One software install a profile (or load a LUT in the graphic card) if this is not to match real colors?
I am puzzled??? Please explain.
Regards,
Antoine
Antoine,
When you see the colours change in Windows, than that's the result of calibration. The videoLUT is adjusted in an effort to make the screen neutral. This is not the result of colour management. It's just adjusting the colours.
The screen profile, however, is used by colour managed applications. It describes the device response of the screen (i.o.w. what the calibration process had to do to make it right).
In your case I'm almost certain the profile is bad.
Rob
Pardon, if I overlooked.
But color difference may occur, if you have "Convert to working RGB" choosen in PS 'Color Settings' when opening untagged sRGB image and PS's Working RGB space is Adobe RGB (or any different from sRGB).
Just an example. Make Adobe RGB as your working space and the policy — 'Convert to working space'.
Load untagged sRGB image. Watch. Then Assign sRGB Profile to the image. See the difference?
Hello, back in april I started this thread. Since then I did more research on this subject and wrote a new article visible there: http://antoine.garric.free.fr/profile_colors_issue/
This is the summary:
After calibrating and profiling my monitor I had problems with some images (see "The problem"). I first though that this was a bug with Photoshop (see "The wrong conclusion"), I know understand that this is a much more complex issue: general users misunderstanding, poor user interface and technology limits (see "The true explaination").
I hope color experts can recommend a solution (see "I found no solution to my problem!") and also help me understand general questions with color management (see "The big question").
First of all, when testing your original jpeg, I don't see your color shift:
Second, when looking closer at your screenshot (assuming it is correct), you'll notice that the red is indeed clipped - but not in the red channel. It is the green channel that is heavily clipped in the low end. IOW, it's clipped towards magenta.
I find it difficult to believe that this is the result of different rendering intents. Everything points to a bad monitor profile.
That said, I've never bothered to play around much with the rendering intents. If it's out of gamut it's out of gamut, and I prefer to deal with it in my own way before the conversion (after having established that it is out of the target space gamut).
D Fosse a écrit :
First of all, when testing your original jpeg, I don't see your color shift:
http://forums.adobe.com/servlet/JiveServlet/showImage/12225/cockpit+sc reenshot.png
Second, when looking closer at your screenshot (assuming it is correct), you'll notice that the red is indeed clipped - but not in the red channel. It is the green channel that is heavily clipped in the low end. IOW, it's clipped towards magenta.
http://forums.adobe.com/servlet/JiveServlet/showImage/12226/comparison +cockpit.png
I find it difficult to believe that this is the result of different rendering intents. Everything points to a bad monitor profile.
That said, I've never bothered to play around much with the rendering intents. If it's out of gamut it's out of gamut, and I prefer to deal with it in my own way before the conversion (after having established that it is out of the target space gamut).
>
Hello thanks for your answer and testing, very useful answer!
D Fosse wrote :
>First of all, when testing your original jpeg, I don't see your color shift:
To me it means you don't have the same problem that I have, you are a lucky man!
>Second, when looking closer at your screenshot (assuming it is correct), you'll notice that the red is indeed
>clipped - but not in the red channel. It is the green channel that is heavily clipped in the low end. IOW, it's
>clipped towards magenta.
My screenshot should be correct, i just pressed the
Good idea to look at the color histogram! I din't try that very useful!
>I find it difficult to believe that this is the result of different rendering intents. Everything points to a bad monitor profile.
I wish you were right. I tried to profile my monitor with 2 differents tools/software, and the profiles are very similar and the issue is the same. I will try a third one (someone should lend me one next week) but I not confident.
Assuming that my profile is correct it would simply mean that the colors of my phosphors are very far from the std sRGB, and the result is this cliping from Photoshop.
>That said, I've never bothered to play around much with the rendering intents. If it's out of gamut it's out of gamut,
>and I prefer to deal with it in my own way before the conversion (after having established that it is out of the target
>space gamut).
How do you know that colors will be out of gammut, What tool do you use?
Again thanks very much
agarric wrote:
How do you know that colors will be out of gammut, What tool do you use?
Well, you look for clipping in the histogram, and look at individual channels to see where the clipping actually occurs. I do this on a copy converted to the target space. Then I decide if the clipping is worth bothering with. It isn't always.
Soft-proofing (Ctrl+Y) is a quicker method, but purely visual.
edited for clarity
I think the OP was entirely correct in his first post - when a monitor is already calibrated _and_ the calibration LUT is already applied (either via the callibrator's own LUT software or by Vista/Win7's LUT loader), colours are wrong, probably because the correction is being applied again (incorrectly) by Photoshop.
Take this scenario:
- We want to author graphics for a non-colour managed application, so we set PS to work in sRGB (this is used by Windows).
- We also have calibrated the monitor for sRGB (in my case using Pantone's Huey).
- The calibration data is loaded into the graphics card LUT (eg. by Windows 7), so the screen is showing sRGB correctly system-wide.
As everything is now set for sRGB, Photoshop _should_ show colours exactly the same as any non-colour managed application - but it doesn't!
I think what's happening is that PS (CS4 in my case) is assuming the monitor calibration data is NOT loaded/applied already, and so applies the calibration conversion a 2nd time! This behaviour would be correct for earlier versions of Windows as they did not load/apply the monitor calibration data, but it's wrong if it's already applied!
The workaround is to use PS's soft proofing, set for 'Monitor RGB' - now the colours finally look correct. But this is the wrong solution - it is supposed to _bypass_ the monitor calibration.
How can we fix this?
I think the OP was entirely correct in his first post - when a monitor is already calibrated _and_ the calibration LUT is already applied (either via the callibrator's own LUT software or by Vista/Win7's LUT loader), colours are wrong, probably because the correction is being applied again (incorrectly) by Photoshop.
No.
Apparently you misunderstand how display profiles work.
The correction LUT just brings the monitor back to some "standard" state -- calibrating it.
The rest of the profile describes the state of the monitor and how colors appear on the monitor - that is what Photoshop uses to correct images for that monitor.
There is no system wide color correction - just a single LUT loaded into the card to bring it to a known state.
There is no, and can be no, double correction in this scenario.
If images look wrong in Photoshop, then the display profile is certainly wrong (not correctly describing the state of the display and how colors appear on it).
As I understand it, the calibration (to sRGB) corrects the monitor so it displays sRGB accuractely (once the correction is applied to the graphic card's LUT). So the 'standard state' in this case is sRGB. I have 4 monitors, all calibrated, all showing colours the same way.
So when PS is in sRGB mode, there is no reason for it to correct its colours again.
And there can certainly be a double-correction, if Photoshop translates the colours before handing them to the graphics card, and then the LUT correction is applied (which is system-wide). Only older versions of Windows didn't actually apply a LUT system-wide, since Vista Windows does this.
EDIT: What you may be missing is that sRGB is Windows' own colour mode. So the calibration devices calibrate to show sRGB accurately (the 'standard state'). So we're not talking about converting between different colours spaces.
The calibration makes sure your monitor is displaying something close to sRGB.
That doesn't mean that applications don't have to do any correction - because what they are displaying may not be sRGB, and they still have to convert TO sRGB for display.
If your display is calibrated to sRGB, and your display profile is sRGB, and your document is sRGB - then there will be no conversion for display.
And there can certainly be a double-correction, if Photoshop translates the colours before handing them to the graphics card, and then the LUT correction is applied (which is system-wide).
No. Again, you don't seem to understand how the calibration and correction are working.
There is no system wide color correction.
Photoshop has to do the color correction itself.
The calibration process just ensures that your display matches the profile you specified. Apps still have to use that profile to convert colors values to appear accurately on your display.
Only older versions of Windows didn't actually apply a LUT system-wide, since Vista Windows does this.
That is incorrect as well. Older and current versions of Windows do the same thing - they load a LUT into the video card to help calibrate the display. No version of Windows does system wide color correction.
Again, there is no chance of double correction here.
If images are not displaying correctly in Photoshop, you can be assured that your display profile is wrong.
I'm afraid you're wrong. I'm a programmer, and I've written my own LUT loader precisely because XP doesn't load LUTs - you can apply correction profiles to each screen but nothing happens (except in colour-management aware applications).
Vista and Windows 7 _do_ now load LUTs for system-wide correction.
XP (and 95/98/ME/2000/NT) could load LUTs - that's how display calibrators worked. It just didn't do so automatically and needed a separate app/driver to do the loading.
The only thing Vista/Win7 added was a built in calibrator and LUT loader -- which still do no system wide color correction.
Again, your understanding of how display calibration and color correction works seems far from complete.
I also suggest you take a minute to read the Photoshop splash screen.
Arrogance is never pretty.
Of course I was talking about an _OS_ LUT loader (if I wrote my own it obviously was possible via 3rd party code).
And that is a system-wide correction, in as much as it corrects all monitors to display sRGB (the native colour space) correctly in all applications.
And your assertion that the calibration only 'sort of' gets you to sRGB is wrong too - think about it, if Photoshop can correct the colours, why can't a cablibration utitlity do the same? Of course they do, which is why Photoshop should leave the colours alone.
Let's put it another way: you said
"If your display is calibrated to sRGB, and your display profile is sRGB, and your document is sRGB - then there will be no conversion for display."
That's exactly the scenario I gave. In this scenario, Photoshop CS4 on Windows 7 (with system-wide correciton LUTs applied by the OS) is messing with the colours _unless_ I choose soft proofing->MonitorRGB.
And that is a system-wide correction, in as much as it corrects all monitors to display sRGB (the native colour space) correctly in all applications.
But that is not what it does.
Again, you don't seem to understand what the LUT does and how the profile is used by color managed apps.
The LUT does not implement color management - it just brings the display back to a known state. You're just loading a 1D LUT, which is a small fraction of the profile for your display.
if Photoshop can correct the colours, why can't a cablibration utitlity do the same?
Ok, you really don't get it.
Photoshop has the colors for a document in colorspace C, and has to correct those colors for display D.
Your LUT is just making sure that as your display drifts, it stays close to the conditions specified in profile D.
But the conversion needed to get from C to D is not a simple 1D LUT -- it is a 3D transformation of the color values via multiple matrices, 1D LUTs and 3D LUTs. (see http://www.color.org/ for details about what profiles contain and how the calculations are performed)
Photoshop is doing exactly what it has to do in order to get you accurate color - because there is nothing automatic about it. The OS does nothing for color correction.
That's exactly the scenario I gave. Photoshop CS4 on Windows 7 (with system-wide correciton LUTs applied by the OS) is messing with the colours _unless_ I choose soft proofing->MonitorRGB.
Which tells us that at some point, you were mistaken.
One of your profiles didn't match, or was incorrect to start with.
There are only 3 ways to get bad color out of Photoshop:
1) Mess up the color yourself
2) Use a buggy video card driver
3) Use a display profile that doesn't accurately represent your display
You clearly haven't read my posts carefully.
Full disclosue - I fully understand LUTs, 3D matrices for colour correction etc. I've been programming graphics for a long time (including graphics card shader stuff).
However I don't intimately know how display profiles work, so see if you agree with this:
- PC monitors are generally designed for sRGB (ignoring wide-gamut displays).
- Calibration utilities produce display profiles with RGB gamma curves to attempt to correct for colour inaccuracies in panels & monitor settings.
- These correction curves are applied system-wide through the use of LUTs on the graphics card (by whatever driver/code).
I have 4 monitors from 3 different vendors, all calibrated with Pantone Huey. They are very close in their colour display now, which suggests to me that the overall colour balance is reasonably close to sRGB.
However, when I view images in Photoshop, set to sRGB space, with an sRGB document, there is a colour shift.
Are you sure that this shift is correct? I don't think this happened with PS CS2 on XP (I can go back later to double-check).
Even if I set PS's colour settings to 'Monitor Colour' (via Colour Settings), this shift persists - the only way to eliminate it is via soft-proofing to 'Monitor RGB'. This simply seems wrong, and if it's not, these colour options are completely unintuitive.
Again, I have read your posts carefully. You have made a lot of mistakes and I've been trying to correct them. Some seem to be mistakes of understanding, and some you just may not have explained what you meant correctly.
BTW - Current displays are leaning toward Adobe RGB more than sRGB. Wide gamut is becoming the standard.
The gamma curves (1D LUTs) can only correct the Tone Response Curve and white point of the display. They cannot correct for primaries that do not match a standard, and few flat panels match sRGB primaries.
I have 4 monitors from 3 different vendors, all calibrated with Pantone Huey. They are very close in their colour display now, which suggests to me that the overall colour balance is reasonably close to sRGB.
No, that just means that they match. That does not mean that they are sRGB. It doesn't even mean that your Huey or the driver software is working correctly.
I have 3 monitors in front of me that match, and all are calibrated to D65 Adobe RGB primaries. And I found out the hard way that Hueys don't work on wide gamut displays.
However, when I view images in Photoshop, set to sRGB space, with an sRGB document, there is a colour shift.
And that tells us that the displays are not exactly sRGB, or that the display profile is incorrect for your display.
For other people, it works just fine. In our testing, and beta testing (with the guys who literally write the books on color), it works just fine.
The only time it fails is when the user configures preferences incorrectly, or has a display profile that doesn't match the display.
Correct - you should not have to use "proof for monitor RGB".
Somewhere, you have made a mistake, or used a profile that is not correct. And your misunderstandings about what should be happening seems to be preventing you from correcting the problem.
OK, let's keep it simple (I will check my XP/CS2 install later to see if there is a difference in behaviour - I never noticed a shift there with the same setup).
I'm just designing UI graphics for a (non-colour managed) application. I want to see the colours of my (calibrated) monitors in PS, without any shift (because that's how the final UI will look on them, and my basic monitor calibrations are a better target for me than any more complex preview of a 'theoretical' sRGB mode).
What Photoshop settings would you suggest to achieve this? You have already agreed that I shouldn't need to soft-proof.
I'll then go and check what is going on. Thanks Chris.
OK, I have figured out what my issue was - it all resulted from my incorrect assumptions about the sRGB space, ie. that calibrated monitors essentially display sRGB correctly. I now see that this was wrong and that Photoshop may indeed need to perform an additional conversion to approximate the sRGB space on such displays. I also found that PS CS2 shows the same behaviour on XP.
In my case, the unexpected colour shift was caused by the document I was working on having an sRGB profile (because of my assumption that sRGB = same as calibrated display). In this scenario, Photoshop will only show the native monitor colours (without requiring Proof->Monitor RGB) if the Colour Settings are set to 'Monitor Colour' or 'No Colour Management' _and_ the document profile is set to 'Dont colour manage this document'.
Not intuitive - you would assume 'No Colour Management' means that Photoshop performs no conversions at all, yet it still does if the source document is tagged with a profile that doesn't match the display.
It's obvious that colour management is still way too complex for the average user - I'm highly technical with a lot of graphics experience, and you could see how I struggled to get it - how is the average PS user going to figure this out? Google around, it's obvious many thoughtful people struggle with it and do bad things like remove all colour profiles just to get the result they want.
I appreciate there are good technical reasons for the complexity, however the way this is all exposed to users needs simplifying. Many users like myself, want to work without colour management in some scenarios, yet they are expected to understand pretty much the entire tangled topic before they can understand which multiple settings are required to achieve it. A few common-scenario wizards would go a long way to make PS more usable out-of-the-box.
Anyway, thanks for steering me in the right direction Chris.
function(){return A.apply(null,[this].concat($A(arguments)))}
function(){return A.apply(null,[this].concat($A(arguments)))}John Joslin wrote:
<sigh>
John, I can only assume you don't understand the concept behind what I wrote. I'll be happy to teach you.
If, on the other hand, your "editorial" comments are intended to show how limited my thinking is as compared to your own perfect understanding of the color world, I can only suggest you try to imagine that someone could actually grasp things beyond even the vast sea of your knowledge, out there where the water runs off the flat surface of the planet... I know it's hard...
-Noel
I don't understand Noel's comment either!
Seems like every few months we go through this exercise, and each time, if I take the time to read it, I learn I don't know JS about color management. This time is no exception! Yet, I have been involved in instrumentation all my professional life as an engineer. Calibration, when you are the source of instruments, is a necessary obsession! And sources of aggravation! At one point, I was taking measurements on infrared lasers, and we had three optical power meters, all calibrated to within 3% by the same lab, yet the spread between them was 10%. 6% should have been max. I fought that for weeks, the solution to which is beyond the scope of this thread, but I just thought it an interesting aside.
So, musing the difference between calibration and profiling, one thought occurred: If the system is properly calibrated (I emphasize system here) why is there a need to profile? Are there non-linearities at certain in-between colors, like RGB is correct to say +/- 1, yet mauve looks wrong? Mauve is supplied as the color but the value read by the meter says wrong by X amount and tells the system it is so? It would seem to me that mauve would have to be corrected(calibrated) so what I see is what I get, which profiling is supposed to do, I presume. Therefore, profiling is calibrating from my pov, which is not correct, from, say, Chris pov. LUT's must then supply information to the system that a particular color is not what is seen, therefore correct the display so it is, which is a calibration in my book.
In all the instrumentation measurements in which I have been involved, subjectivity of the observer is identified and compensated, such as parallax. Color, on the other hand, is rife with subjectivity (my SO, who is a colorist of the old school of color mixing and I have huge arguments about what is red, or blue.(Her primaries are red, yellow and blue)), in which only wavelength overcomes. But what is the wavelength of magenta? And so on.
Yep, I don't understand color management! Thank goodness that my Eye1 and software is actually doing the job, and that if what I see is not what's there, I look to my devices and software for the fix. My print output is the final arbiter.
North America
Europe, Middle East and Africa
Asia Pacific