Mike Chaney's Tech Corner

Hi,

I've been investigating some color management issues that I had assumed were with my printer, but it looks like part of the problem is with QImage. I know that color management is easy to mess up, so bear with my and I'll try and demonstrate that I've been reasonably thorough before suggesting a QImage problem.

I wanted some standard test targets to experiment with, so after some searching I downloaded the following synthetic test charts from http://www.babelcolor.com/main_level/ColorChecker.htm

ColorChecker_Lab_from_Avg.tif
ColorChecker_ProPhoto_from_Avg_16bit.tif
ColorChecker_AdobeRGB_from_Avg_16bit.tif

These are slightly odd in that they are not tagged with their respective color profiles, so on loading them into photoshop they each look visibly different. I assigned the appropriate color space to each of the rgb files (all three then looked the same on screen) and saved them. I then printed all three via QImage onto a variety of different papers. In QImage the color spaces are correctly identified.
What was immediately obvious was that the Lab version was very different to the other two versions. I also found that switching between perceptual and absolute colorimetric caused the proPhoto version to take on a marked warm shift on the grey scale. In previous versions the AdobeRGB version took on a cool shift, but this seems to have gone after updating to 2010.100

I realised I can show this by printing to a jpeg, here's an example with the three versions of the chart, using AC rendering intent:
(http://farm3.static.flickr.com/2694/4036414250_d21fda747c.jpg)

Perceptual looks better, but the Lab is still very different:
(http://farm4.static.flickr.com/3536/4036414174_dd8d0d2ba0.jpg)

I then tried doing the same thing printing from Lightroom, and all three versions rendered very similarly (on the print, I don't think I could tell them apart). Lightroom can also print to file, so here is the lightroom version:
(http://farm3.static.flickr.com/2580/4036414062_4bc3993237_o.jpg)

These test ouputs are on flickr if you want to look in more detail:
http://www.flickr.com/photos/74697159@N00/sets/72157622518862305/ (http://www.flickr.com/photos/74697159@N00/sets/72157622518862305/)


I can only claim a partial understanding of the issues of color spaces, but it looks to me like there is something odd going on here!

Regards,
James.

That is extremely unlikely; there would have been numerous complaints by now if there was  :o

Not really, no calibration target has a profile associated with it.

Assigning in PS dose not necessarily mean that other software will read that information. The sure way is to embed the profile.
How did you check that? Mouse over the thumbnail and the colour space will show in the exifbar at the bottom of the screen, or right click and Display Image Info. That will tell you if there's an embedded profile.
To be sure about the relevant profiles in Qimage, you can use a filter to override whatever is associated with the image, that way you'll be sure Qimage is using the one you want.
Terry.

James,
I've just repeated your checks with sRGB, Adobe RGB, AppleRGB and ProPhoto colour spaces. I did not know how to deal wit the Lab one in terms of an icc so ignored it.
I use a filter to "override" the icc and printed to file in Q so the image you see below is sRGB

See attached, there's no difference in appearance on any of them.
Your problems lie outside of Qimage I think, or you are not using it correctly  :o  ;)
Terry.

Terry,

Many thanks for looking into this for me. I agree, your results look fine. Comparing your output with mine, it looks like you used perceptual rendering intent. I suspect that my problems are due to a lack of understanding of the Lab format and the use of the AC intent, but since I don't normally use either it's really a non-issue for me. I'll do another set of tests with the file formats and intent that I actually use in practice, and hopefully that will confirm that all is well!

Thank you again,
James.

It was actually Relative Colormetric.
best of luck with sorting this out  :)
Terry

Interesting. I removed the images from flickr as I didn't want to leave an incorrect impression of QImage's capabilities in a public place when in all likelihood the problem is user error. I'm guessing my original post will show broken images at some point (doesn't seem to have yet), so I'll attach the main image of interest in case anyone else wants to play with Lab or AC intent. (I had to resize it to meet the max size limit)

I repeated the tests using the color spaces and intents that I actually use, and everything was fine. Panic over. I'll put the odd cases down to a lack of understanding of Lab and the absolute intent on my part.

All the best,
James.

Hi James,
In PS, it easy to 'assign' a colour space, when I expect you may really need to 'convert' to the colour space. I'm not sure/not checked, :-[  if you save it as a psd if the file  then contains the assigned colour space, but it does not if saved as jpeg, for example. This tends to commonly give drab/reddish images if you're working in Adobe rgb, say, but display in srgb. It'll probably look OK on your monitor, in ps, but not if others view it on the web.

Best wishes,

Ray