[argyllcms] Re: How can I tell if I'm using my profile? (and other questions)

In my case I believe that I ran 'dispwin profile.icc' without the -I option. Does this load the profile or does it only load the calibration contained in the profile? I think that you have to restart firefox to get changes in the profile to be recognized. I have the color management extension installed in firefox and there is a box where I can specify the profile filename; when I specified the color swapped profile and then restarted firefox I did observe the swapped colors used in firefox. (In fact, they were used throughout the application, for menus, backgrounds, etc., so the change was very obvious.) But if it's not honoring the properly installed profile perhaps we need to submit a bug report. I was displaying my images with xv which I'd be willing to bet does not do color management. Just to confirm that I understand properly, if I were to calibrate my display to gamma=2.0, say, it wouldn't matter if all the applications were color managed because they'd look at the colorspace for the images and make the appropriate translations. The only reason it matters is because of the non-color managed apps. Is that right? But is that really true? If I only have 8 bits to work with and all of my images are really in sRBG (say) then if I calibrate to gamma=2.0 then I need a second conversion to get by sRGB data mapped to my display. Wouldn't it be optimal in this case to instead calibrate the display to sRGB so that I only have one layer of mapping (in 8 bits) to display the data there? In other words, I should get less banding and artifacts when viewing the sRGB source data if I calibrate the display to sRGB even though this is farther from the display's native response. It seems like I should calibrate to the expected color space of my data. But if I have no idea what the data's colorspace is, or data comes in many different color spaces, then I should calibrate to the monitor's native response. (Or not calibrate at all? Only profile?) Can anybody direct me to a good image to use to test for banding? 
Florian Höch wrote:
Interesting. I was not aware that Firefox does not honor the display profile under Linux (under Mac OS X and Windows it works for me), but a quick test on my own Linux box also didn't show expected results :(
Regarding your gamma question, I think in your case you can choose what suits you best (a reason to try and approximate the display's native gamma response is maybe to reduce banding, if that is an issue). The chosen gamma value when calibrating makes a difference for non-colormanaged cases (e.g. the desktop and most other software). If you see differences in your image viewer between different gamma calibrations, it could mean your images do not contain profiles (although a lot of consumer-grade digital cameras output sRGB or atleast tag as sRGB), or the used image viewer does not use them, does not honor the display profile, or does not use any profiles, maybe because it is not colormanaged. BTW, regardless of the chosen gamma, calibration tries to correct the gray balance, so a calibrated display should yield better grayscales (neutrality with respect to the whitepoint).
Regarding the linux ICC-aware image viewer, I think GIMP works after some configuration but I haven't tried it. If anybody knows a suitable viewer, I would also be interested. 

Regards,

Florian

Adrian Mariano schrieb:

Florian Höch wrote:

Adrian Mariano wrote:
I ran dispcal with the -o option so as to produce a profile as output. I loaded the profile using dispwin under linux. Now I go into firefox 3 and color management is turned on. Does firefox find the profile that was loaded using dispwin?
You can test if Firefox 3 uses your installed display profile by temporarily installing a profile with swapped colors, like this one: 
http://hoech.net/files/BRG.icc
Close your browser, install the above profile using dispwin -I BRG.icc
Open your browser and visit http://www.color.org/version4html.xalter
The big image should have magenta-red sky and green mountains, and look like the topmost small image, if the display profile is used.
Thanks. That image clearly revealed that firefox was NOT using the profile that I thought I had installed with dispwin. (Now I'm looking back at what you wrote and I'm not sure I gave '-I' when I ran dispwin, so I'll test that and see if the behavior is different.)
I made the above profile using sRGB, but I noticed the suggestion in the documentation that perhaps a gamma of 2.4 would be a better choice? I think the last time I ran a calibration I specified a temperature of 6500 K, but this time I forgot that option, which apparently means the monitor's native white point is used. Am I better off using the native white point?
Again, depends. Most (LCD) monitors offer white point control only by adjusting their internal LUT curves to reach the target value, thus sacrificing some of the 256 possible levels per channel and introducing the danger of visible banding in gradients (at 8 bit, some more expensive displays also haver 10 or 12 bit internal LUTs). So yes, it might be better to use the native white point. A gamma of 2.4 is a good starting point, you could also use dispcal -R to get a measurement of your uncalibrated displays approximate gamma response and then use that.
Really? My interest in calibration was driven by being baffled by my uncalibrated display when I looked at gamma test patterns because the display seemed to be so distant from the normal expectations, so I wouldn't think I should TRY to make the display look like it is uncalibrated. (It helped that a friend gave me her colorimeter because the latest version of Windows didn't support it any more!) Here is the result of dispcal -R: 

Uncalibrated response:
Black level = 0.29 cd/m^2
White level = 168.02 cd/m^2
Aprox. gamma = 1.97
Contrast ratio = 579:1
White chromaticity coordinates 0.3146, 0.3400
White    Correlated Color Temperature = 6333K, DE to locus =  9.8
White Correlated Daylight Temperature = 6330K, DE to locus =  6.3
White        Visual Color Temperature = 5996K, DE to locus =  9.4
White     Visual Daylight Temperature = 6128K, DE to locus =  6.1
Effective LUT entry depth seems to be 8 bits
The instrument can be removed from the screen.
So you would have me set my gamma at 2.0 so as to match the uncalibrated display? I generated two profiles with the natural white point (which seems reasonably close to 6500K) and with gamma 2.4 and with sRGB. I viewed images taken with my digital camera using the calibrations from these two profiles (in software that I'm pretty sure is not color managed) and we looked at them and tried to decide which looked better, and it seemed like the sRBG version looked kind of washed out compared to the gamma 2.4 version. My wife thought the gamma=2.4 pictures looked clearly superior. (I suppose it could be that we take overexposed pictures and the 2.4 gamma is correcting a defect rather than properly rendering them.) I also found this site which has some interesting examples where you can mouse over an image and it switches between the tagged and untagged versions. http://www.gballard.net/psd/go_live_page_profile/embeddedJPEGprofiles.html
The author of this page strongly insists that you should calibrate your display to sRGB (which was what I thought was the right thing to do before I saw the suggestion of gamma=2.4 in the dispcal manual). Can anybody suggest software for linux which can display jpeg images and can respect, display, and edit color tags in the images? (Does my canon digital camera include some kind of color space information in the files? We tried viewing the images in firefox and outside and it seemed that there was a slight difference, which suggests the answer is yes.)

Other related posts: