ColorChecker Profile Results: Questions

[nilo]

[UPDATE: Problem solved: Schewe says that I just imagine the difference in color on the floor.]

I have these two cameras that have to get close for the colors, but the result I get with the passport are not usable.

Here is the setup:
I took two pictures of the ColorChecker 24. One with a H4D-40 and one with a 5DMk2. Both cameras where mounted on the same studio stand at the same spot, one after the other, same studio light. Identical exposure settings measured with a light meter. Both images preprocessed with the ColorChecker Passport plug-in. I applied the respective profiles, white balanced on the second gray patch from the left. Finally, I slightly adjusted for the differences in exposure (+- 0.10), did the white balance  again and cropped the 2:3 5D2 image only on the sides to match the 6:4.5 of the Hassy.

Why are the colors SO different? What did I do wrong?

sincerely
nino

[rasworth]

IMO a better tool for matching two cameras is the Adobe DNG Profile Editor,

http://labs.adobe.com/wiki/index.php/DNG_Profiles

It has more capability to "bend" the color checker colors than the Passport.

Richard Southworth

[nilo]

IMO a better tool for matching two cameras is the Adobe DNG Profile Editor,

http://labs.adobe.com/wiki/index.php/DNG_Profiles

It has more capability to "bend" the color checker colors than the Passport.

Richard Southworth

So you are talking visual matching, like one would do without the colorchecker software, right? I did this in ACR/LR directly. As the profile is anyway not portable, no need to make a "profile" in this software beforehand. Also, the colors are so far apart, that I could just start making a "profile" by hand. I don't see what the Passport or the DNG Profile Editor are doing here. The advertisement said that it would help to bring different cameras in sync...

[rasworth]

No, not visual, separate program like Passport.  It lays a LUT (look up table) on top of the matrix conversion, therefore in my experience can do a better job of matching two cameras.  Again, it creates individual DNG camera profiles, just like Passport.

Richard Southworth

[nilo]

No, not visual, separate program like Passport.  It lays a LUT (look up table) on top of the matrix conversion, therefore in my experience can do a better job of matching two cameras.  Again, it creates individual DNG camera profiles, just like Passport.

Richard Southworth

I used the DNG Editor, but the tweaking of the results from the two cameras, in order to make a better match, is by visual assessment. What I wanted to say, is that it appears to me that I can get to the same result by tweaking the colors directly in LR/ACR and paste those adjustments to the corresponding image. No need for any passport and/or DNG editor, as their results are anyway far apart...
This is how I see it now, from experience (see attachments posted above). Maybe someone can explain me where I'm wrong. One answer I could think of is, that the passport etc helps you get close easily and quickly. The rest you'll have to do by hand, through visual assessment via the DNG editor.

[rasworth]

Sorry for the confusion, use the DNG PE in its automatic mode, i.e. "Chart".  Do this for both cameras with the targets you've already generated.  Use the profiles as generated with no manual tweaking.

Richard Southworth

[nilo]

Sorry for the confusion, use the DNG PE in its automatic mode, i.e. "Chart".  Do this for both cameras with the targets you've already generated.  Use the profiles as generated with no manual tweaking.

Richard Southworth

The result of which you can see in the two attachments in my initial post. I applied them back to that original target from which the profiles were created. The question is now, how can it be that there is such a huge difference, when
- It is one and the same scene
- the same lighting situation
- on target shot after the other
- controlled lighting
the rest of the procedure is described in my original post:
 

I have these two cameras that have to get close for the colors, but the result I get with the passport are not usable.

Here is the setup:
I took two pictures of the ColorChecker 24. One with a H4D-40 and one with a 5DMk2. Both cameras where mounted on the same studio stand at the same spot, one after the other, same studio light. Identical exposure settings measured with a light meter. Both images preprocessed with the ColorChecker Passport plug-in. I applied the respective profiles, white balanced on the second gray patch from the left. Finally, I slightly adjusted for the differences in exposure (+- 0.10), did the white balance  again and cropped the 2:3 5D2 image only on the sides to match the 6:4.5 of the Hassy.

Why are the colors SO different? What did I do wrong?

[rasworth]

My interpretation of your original post is you used the XRite Passport software, not the Adobe DNG Profile Editor software for which I gave you the link.  I apologize if I misunderstood, but we are talking about two different software packages used to create DNG camera profiles compatible with Lightroom and ACR/Photoshop.

Richard Southworth

[rasworth]

I downloaded your two images, loaded into Photoshop and assigned the sRGB space to each.  I then balanced the exposure using the neutral patches (they were about .1 stop different) and visually/numerically compared the two images.  There is very little difference, I am now confused as to why you were disappointed.

Richard Southworth

[madmanchan]

It is possible the Passport software does not try to build profiles that perform exact colorimetric matching of the chart patches.  I think it more likely that it does a preferred/perceptual rendering using the input color patches as a guide, rather than the final goal.  (The thinking being that ultimately you will be photographing scenes with other types of objects, rather than charts.)  This means that you can feed chart images from different cameras or different situations and not get exact colorimetric matches when applying the profiles.

I should also caution you that even if you were able to get exact colorimetric matches between the 5D2 and H4D-40 for the chart patches, in practice for real scenes you would not be able to get consistent matches. This is because the sensors' spectral sensitivities for the 5D2 and H4D-40 are quite different.

Eric

[nilo]

I downloaded your two images, loaded into Photoshop and assigned the sRGB space to each.  I then balanced the exposure using the neutral patches (they were about .1 stop different) and visually/numerically compared the two images.  There is very little difference, I am now confused as to why you were disappointed.

Richard Southworth

I am sorry for not being clear, Richard! I did indeed use the Passport software, but was in the believe that it is basically the same as the DNG PE. I used both and couldn't notice any difference as to how the "profile" comes out. I didn't measure anything though. These two images are the final result, the "profiles" have been added to the image from which they have been created, nothing more (EDIT: I did the white balance again, after applying the newly created profile). I didn't measure the ColorChecker afterwards, I just had a look on the overall different "color cast". Just look at the paper on the floor, or the toys! The two cameras produce so different colors, that it is not possible to use them together like this. As we said, it would need further visual tweaking either in LR/ACR or DNG PE to make the two come closer. That's fairly easy, and is what I had to do. But the Passport, for me, does not work like advertised, meaning, bringing two cameras together. So I am trying to understand what I got wrong. Like things stand for now, there is no difference in using the x-rite colorchecker software or not, 'cause you'll have to tweak anyway. Tweaking right away just safes time. Or maybe not, because it helps to get the values closer, and tweak from there. But then everyone should/must use the DNG Profile Editor together with this product. In fact they should ship together... I hope I could explain myself a bit better.

As a side note: Why did I try the Colorchecker software(s) at all? Because I regularly want to check results against my actual workflow, which is via ICC profiles. That way, IMO the two cameras come easily close enough to be used together. provided you take maximum care on how you shoot the target (one light only in the black) and which target (gloss, natural pigments and a lot of patches if for LUT based profile).

[Schewe]

I think, even though the setup is not scientific, the conditions a quite good, and even a lot better than standard usage, and should therefore, IMO, give a lot better results.

Sorry bud...but when you first posted, I looked at the two images and truth be told, had no clue what you were/are talking about. Honestly when I looked at the two images above, I'm thinking–this is a trick question, right? Cause both images as posted looked for all the potential variables–as close to the same as anybody could expect. Seriously, what are you expecting? The colors of the two ColorChecker shots are, to my eyes, really darn close.

So, exactly what are you expecting? Cause I gotta tell ya, those two shots you posted are, well, pretty darn close. Not at all sure what you are expecting, but this example has gotta be close to "optimal"...unless you know something I don't (which isn't really likely if you know what I mean :~).

[Schewe]

sorry bud if you can't see it.

No, seriously, aside from a slight exposure adjustment, what I see in my web browser (Safari 5.1 (6534.50)) is close enough to be totally negligible...sorry, I'm with rasworth...much ado about nothing. Maybe you should double check your settings (just saying').

[Schewe]

Check your monitor settings.

Uh huh...really? Do you want to go there? My display (it ain't a "monitor" doooode) is just fine. Wanna come over to my house and see? Since Safari is mildly color management savvy (and even though I'm look at the images on a NEC PA242W which is, of course profiled) I really don't see much difference at all between the two images posted. Is that really my fault?

Again, what EXACTLY are you asking...what EXACTLY are you expecting?

[Schewe]

I updated my original question. Thank you.

So...never mind? I'm still wondering about what you thought you were seeing and what you were expecting. While you have retroactively modified your OP, I still am at a loss what you thought you were seeing and what you were expecting?

[Schewe]

Not everybody has to be mean ;-)

Just to be perfectly clear, I didn't approach "mean" in any way...that's baggage YOU are bringing...I'm actually curious (not in the least bit "mean", yet–I can get "mean" if ya want)...

Really, it's my experience that a properly made DNG profile–regardless of the original camera–should be pretty close...if you aren't seeing that, it would be useful to know why?

[digitaldog]

I would also agree that the two are extremely close visually on a browser, I do see a slight difference in gray balance on the back wall. Using the Apple DigitalColor Meter as the image appears in Safari, the upper image is about 1-2 units higher in blue.