How to ICC profile a camera? (no DNG)

[Chris_Brown]

Before you import the TIFF file into Basiccolor, open it in Photoshop and strip out the ICC profile.

Edit > Assign Profile > "Don't color manage this document".

Then proceed to Basiccolor.

[nilo]

I tried it immediately, but the result is the same, way too dark.
 

[nilo]

The darkness in fact is not critical. I mean, I could edit the icc for that. Just what about the crushed shadows? Maybe that realy stems from the CC24 lack of black patches? The question could therefore be SG vs basICColor target

regards

[nilo]

Chris,
here is a little piece of it where you can see the horrible posterization:

what is causing this? the cc24?

[Bart_van_der_Wolf]

Chris,
here is a little piece of it where you can see the horrible posterization:

what is causing this? the cc24?

Possibly, and 24 patches is not enough for a good Camera profile, that's why the Colorchecker SG has 140.
I've made camera profiles in the past with the i1Match software that comes with the EyeOne Pro, but for better profiles you'll probably need something like ProfileMaker Platinum. I'm waiting for the new X-rite i1Publish Pro (looks like they are not going to make 2010Q4 release time), to see how the integration is for camera profile generation and tweaking.

Cheers,
Bart

[nilo]

Bart,

thank you for tuning in!

The new x-rite i1 solutions will not include ICC profiling for cameras according to Andrew Rodney , and according to what can be understood from their web page (http://www.luminous-landscape.com/forum/index.php?topic=49647.msg409494#msg409494).

ProfileMaker didn't do better on this exact same file. Here is a comparison of the renditions of PM vs basICColor vs ArgyllCMS with the same file once linear and once standard curve in Capture One Pro 6.0.1:  http://www.luminous-landscape.com/forum/index.php?topic=49793.msg410438#msg410438

On CC24 vs SG, here is a paper which shows that there is not so much difference between them as one might think: http://www.betterlight.com/downloads/conference06_notes/color_Accuracy-ppt.pdf

But, like you, I also tend do believe that this is the problem. If so the basICColor target should do a much better job than the SG (and cost also much much more).

 

regards

[nilo]

The best results should be possible with the basICColor target, but processed in ArgyllCMS.

[Bart_van_der_Wolf]

Bart,

thank you for tuning in!

The new x-rite i1 solutions will not include ICC profiling for cameras according to Andrew Rodney , and according to what can be understood from their web page (http://www.luminous-landscape.com/forum/index.php?topic=49647.msg409494#msg409494).

Yes, it seems like it won't, initially anyway. We'll see how that pans out. I'm still sceptical about just using 24 patches though, e.g. based on printer profiles based on many more patches which are significantly more successful in nailing the output characteristics.

ProfileMaker didn't do better on this exact same file. Here is a comparison of the renditions of PM vs basICColor vs ArgyllCMS with the same file once linear and once standard curve in Capture One Pro 6.0.1:  http://www.luminous-landscape.com/forum/index.php?topic=49793.msg410438#msg410438

It's hard to judge someone else's workflow, especially since there usually is a learning curve involved in mastering the various software solutions. I haven't tried generating profiles with those pakages yet, so I can't help with that.

On CC24 vs SG, here is a paper which shows that there is not so much difference between them as one might think: http://www.betterlight.com/downloads/conference06_notes/color_Accuracy-ppt.pdf

However, do note that the 24 patches in the center of the SG, are not the same as the ones on the 24-patch CCs! They look the same, but require different reference data for profiling.

But, like you, I also tend do believe that this is the problem. If so the basICColor target should do a much better job than the SG (and cost also much much more).

Assuming one gets a good reference data file, and the profile making software does a good job with it. Profile making involves a lot of compromises, interpolation, and choices. There are also many variables involved that can spoil the party. One needs to work very structured and accurate, and then still tweak the result to one's liking.

Also note that you can use the Capture One Pro's general camera profile for your camera, and with C1 save an adjusted ICC profile.

Cheers,
Bart

[Chris_Brown]

what is causing this? the cc24?

To me, this looks a gamma issue. That's what I would investigate, and I'd look very closely at Basiccolor. I have not used Basiccolor before, so I don't know how it treats a TIFF file from a raw capture (I use ColorEyes). Have you contacted them for support?

I also don't know how the number of patches affects a profile (I would think that more is better, but I've never seen capture targets with nearly as many patches as print/output targets).

Off the Wall Suggestion:
Use your new ColorChecker to make custom DNG camera profiles using Adobe's DNG Profile editing software. Use the two-illuminant method, too. You have spent so much time without getting any results, you might as well try this method. I've made DNG profiles for my cameras and get very good results. This would require you to use Adobe Camera Raw, though.

[Czornyj]

Anybody?

where should I turn to to find some help and answers?

ArgyllCMS mailing list - invaluable source of any information you could (or even couldn't) imagine. There's a lot of helpful participants and - most of all - tireless, omniscient Graeme Gill.

But I'd also bet it's only a matter of CC24, and huge colorimetric distance between its paches. CC SG should give better validation results, as I belive.

[nilo]

[...]
It's hard to judge someone else's workflow, especially since there usually is a learning curve involved in mastering the various software solutions. I haven't tried generating profiles with those pakages yet, so I can't help with that.

What software/hardware do you use? I included the shot of the target with spot color readouts for checking the evenness of illumination in that same thread, as well as all the setting details http://www.luminous-landscape.com/forum/index.php?topic=49793.msg410398#msg410398 To be fair in the comparison, I only used the standard settings and no tweaking.

[...]
Also note that you can use the Capture One Pro's general camera profile for your camera, and with C1 save an adjusted ICC profile.

as I said, this approach is a bit like the DNG Profile Editor, quite useless in a lot of situations, because, while you tweak here you untweak there. Just think of a portrait shot including some orange, red, pink... product!

[nilo]

To me, this looks a gamma issue.

I also though to check this more thoroughly. But I don't understand why the standard procedures to not give a somewhat usable result. I still think that I do something wrong.

That's what I would investigate, and I'd look very closely at Basiccolor.

why?

[...] (I use ColorEyes). Have you contacted them for support?

waiting for an answer

I also don't know how the number of patches affects a profile (I would think that more is better, but I've never seen capture targets with nearly as many patches as print/output targets).

and why shouldn't a standard IT8 chart do marvels? It has a lot of patches and is glossy too (I tried it -no reflections here- and got bad results)

Off the Wall Suggestion:
Use your new ColorChecker to make custom DNG camera profiles using Adobe's DNG Profile editing software. Use the two-illuminant method, too. You have spent so much time without getting any results, you might as well try this method. I've made DNG profiles for my cameras and get very good results. This would require you to use Adobe Camera Raw, though.

Yes, I am using the x-rite/Adobe way too, for some jobs, but I largely prefer the 5D2, the HDH40 and the occasional Leafs through C1, particularly for portraits and landscape. For  reproduction I had to resort to CameraRaw and DNG "profiles"

[nilo]

Thank you Czornyi for you reply,

ArgyllCMS mailing list - invaluable source of any information you could (or even couldn't) imagine. There's a lot of helpful participants and - most of all - tireless, omniscient Graeme Gill.

I follow the list now, and have a try

But I'd also bet it's only a matter of CC24, and huge colorimetric distance between its paches. CC SG should give better validation results, as I belive.

As I said, I too believe that, but x-rite/Adobe goes with the cc24!? Also, did you see this: http://www.betterlight.com/downloads/conference06_notes/color_Accuracy-ppt.pdf
I'll maybe write them...

[nilo]

After a few more tests, it seams clear that the crushed shadows, the posterization, is due to the nature of the cc24.

I tried a simple IT8 chart. There a lot of patches, more than on the CC SG. Also the IT8 is glossy for rich saturated colors.

Results are very impressive with Monaco EZcolor, a scanner software! I had to use this software, because it was the only software that accepted the .mrf reference file provided for this target by x-rite.

Contrary to what is repeated everywhere, the results are very portable. Compared to the generic Canon 5D Mrk2 ICC profile, this one gives far better results indoor and outdoor in a variety of lighting conditions! (Due to the use of this basic and cheap scanner software, I could get a very neutral profile, ideal for reproduction.)

The assertions in this instructive LL video proof to be accurate indeed http://www.luminous-landscape.com/reviews/accessories/coloreyes-commercial.shtml The key to success is the way you shoot the target: One light only @45 degrees and at a great distance from target, cutter cards to further even out the illumination, Black environment to reduce reflection and color casts, perfect WB, spot on exposure for the white patch, a glossy target and an exact reference file for this particular unit.

regards

EDIT: The results are similar to the generic CaptureOne profile, but the differences exist! The main being "accuracy" in the reds, but don't forget to have a look at the green too. The new IT8/EZcolor profile is more linear and therefore very suitable for reproduction. It feels like an expansion of the gamut of the camera.  

[Chris_Brown]

After a few more tests, it seams clear that the crushed shadows, the posterization, is due to the nature of the cc24.

I tried a simple IT8 chart. There a lot of patches, more than on the CC SG. Also the IT8 is glossy for rich saturated colors.

It looks like Basiccolor is the culprit. Did you use the IT8 target with that software?

[nilo]

It looks like Basiccolor is the culprit. Did you use the IT8 target with that software?

Well, I didn't want to report negative things again, so I choose to focus on good news ;-) One thing is that Monaco Ezcolor does not come with a .txt reference file for this IT8 target, because, guess what, x-rite does not supply one, only ".mrf" [EDIT: se note below]. I therefore believe, that similar results should be obtainable from different software with a target made by a different manufacturer.

I don't mind the linear/flat output of the scanner software, on the contrary. I prefer to edit the ICC profile later myself and not let the software decide. I than save the edited profile as a new ICC profile. This is possible with ProfileMaker5 et al. The latter would also offer presets for such edits (which on can further tweak) for various basic profiles: generic, repro, product, portrait, outdoor, b&w...

In the light of this result, the next step for a further gain in quality, if such a thing is possible, should be the HutchColor target. I can't find any better than that, certainly I do see no advantage for the CCSG, do you? The skin tone patches, should also be addressed by the large amount of patches on the HutchColor target. Also,the CCSG is only semi-gloss, which seems like a compromise for practical reasons. Lastly, the argument that the CCSG contains the CC24 is a non-starter for me, because, for one, who doesn't have one, but mainly because the measured values are NOT the ones of the CC24 (see for example this http://cea4p.free.fr/IMG/pdf/Compa_SG_DC.pdf)

regards

P.-S. Once more I would like to emphasize how practically valid I found the arguments by "Integrated color -ColorEyes" to be http://www.integrated-color.com/cecamera/ The profile is very portable and NOT noticeably SCENE DEPENDENT. A profile for every lens and situation would maybe be even better, but , as mentioned in the video referenced above, other factors will the results a grate deal more than this.

EDIT: meanwhile I found that to get the *.CIE data out of the *.mfr (Monaco Reference File), one just has to rename the *.mrf to *.zip, than unzip, et viola the *.cie! Just rename the unzipped file to .txt.

[nilo]

I now  made a DNG profile out of the carefully shot CC24 file. I also made some quick shots of the CC24 in different studio lighting setups. When I apply them in Camera Raw I can't see any difference between them!

EDIT: but definitely differences with "Adobe standard" and huge differences with the Adobe Canon profile imitations "Standard", "Faithful", "Neutral"..., mainly in hue.

[Chris_Brown]

I now  made a DNG profile out of the carefully shot CC24 file. I also made some quick shots of the CC24 in different studio lighting setups. When I apply them in Camera Raw I can't see any difference between them!

Did you use a dual-color temperature setup?

[nilo]

Did you use a dual-color temperature setup?

I don't understand why that should make anything better, regarding the carefully shot CC24, or for any studio shoot. What too temps should that be?

[Chris_Brown]

I don't understand why that should make anything better, regarding the carefully shot CC24, or for any studio shoot. What too temps should that be?

Most digital camera sensors respond differently under different illuminants (e.g., switching from daylight to tungsten). DNG camera profiles address these differences by allowing color adjustments to be specified separately for two different illuminants (usually Illuminants A and D65). See info here.