4
\$\begingroup\$

It's common in my industry for reference/archival photographs to be taken with a combined scale, grey, and colour checker card (often with rectification targets too). In most cases, the standard is a four-panel colour checker. A common card people use is this, or something like it. Photograph of a pocket reference scale - this version is credit card sized, with an 8 centimetre scale, a 20 millimetre scale, a three-panel grey card, four colour reference panels, and a rectification target. The reference colours are red, yellow, green, and blue.

I have been reviewing a collection of unprocessed RAW photos at my current company, and they have cards like this. I've discovered, however, that no one at the company knows how to process RAW photos, and they just use the camera JPEG outputs, so I am working on processing the backlog of RAW photos. I am used to using a 24 or 48 colour checker, but have not used four-colour checkers before, and I cannot find any information on using them for colour correction.

In particular, I need to know how to make use of these in Darktable (v4.2.1). I have the CMYK values of the cards used, and have converted them to the RGB and hex values. But I cannot work out a way to calibrate the colour using these cards other than by eye.

Can anyone suggest a workflow for this scenario?

\$\endgroup\$
3
  • \$\begingroup\$ Can't you just use the white balance's eyedropper on the gray patch to get accurate WB? \$\endgroup\$
    – dandavis
    Commented Sep 15, 2023 at 21:52
  • 1
    \$\begingroup\$ @dandavis that's for white balance correction. I'm interested in colour calibration, which is what colour checkers are for. See the Calibrite ColorChecker Passport for an example of a 24-panel colour calibration unit. \$\endgroup\$ Commented Sep 17, 2023 at 12:24
  • \$\begingroup\$ I don't see how the color would be off if the white balance is good. Usually calibration is applied to outputs like monitors and printers, not photos. If you want realistic color, toe in the levels to paper white and black, blur 10px to kill noise, then check that the color squares are reflective of the pure rgb values. If there is a cast, use the color correct to bump each channel up/down a notch or three until the eyedropper rgb number match what you expect from the card. Then you can apply the same corrections to each photo in the batch/session. \$\endgroup\$
    – dandavis
    Commented Sep 18, 2023 at 23:39

4 Answers 4

2
\$\begingroup\$

Answering my own question here, as there has been some very helpful input, but I think this might give a more accurate answer to the specific question. There was also extensive dialogue around the same question over in the Darktable forum of pixls.us, which contributed a great deal to exploring this question and colour theory.

In short, the card in question, and similar four-panel cards, are essentially a simplified version of the Kodak colour separation guide and similar tools. This means their function is as a (slightly old-fashioned) printing calibration tool; they are intended to allow you to verify the colour properties of a printed copy of a photo and, if required, make corresponding adjustments. In scientific and research contexts, they're also a useful quick reference for checking that an image is true colour or eyeballing any discolouration.

As for using it to correct white balance, yes it is suitable for that purpose, and it does deliver a drastically improved result - as Dan mentioned - but as explored here there are still significant advantages to using a proper colour calibration solution. Four-panel cards, however, are not that solution.

Speaking of colour calibration solutions, as Zeus points out (better than I can summarise) there is a lot of complex colour science underlying this deceitfully complex question, but unless you want to calibrate by eye (and I thoroughly enjoy doing manual colour processing on my non-academic photography) you're basically tied to solutions like Calibrite's ColorChecker, which account for every link of the chain. As they summed up super neatly:

Manufacturers of the calibrated cards like ColorChecker have instructions (usually in the software) similar to "this cell, under a given lighting conditions, have such-and-such chromaticity. When displayed in sRGB, it should have such-and-such relationship between R/G/B."

I have spotted in my searches a few cards that might take a similar role to a ColorChecker, like this awesome scale card for Australian rock art (below), but given the general lack of information on this web page, my understanding is that cards like this are actually intended for visual Munsell colour assessments - essentially as a context-specific pocket alternative to carrying a full Munsell chart - rather than for any form of digital image correction. If I'm reading their webpage correctly, it's another case of relying on correction by eye if you want to use this for digital image correction.

enter image description here

It seems like there's really no alternative to the immensely robust colour science behind high quality 24- and 48-panel checkers. So to answer my own question, no - there is no workflow to retroactively calibrate these images based on the included reference card other than manual correction.

\$\endgroup\$
1
\$\begingroup\$

I'm not sure I'll be able to help you anyway, but what are the subjects of these photos? What's the purpose of the photographs?

But, to answer your question directly, I don't think you can use any colour card other than the SpyderChecker or ColorChecker cards built into the app. I can't see any option to add more unfortunately. You're reduced to making adjustments until those colours match the colours they should match.

However Dan's right that simply hitting the grey should get it 99% there. Because grey has no colour (or very little!), it can be used to uniformly add/subtract from RGB values to make that patch grey. Do the same to the rest of the picture and it's pretty much correct.

There's a line in the manual:

"Users are discouraged from obtaining cheap, off-brand, color targets as color constancy between batches cannot possibly be asserted at such prices. Inaccurate color checkers will only defeat the purpose of color calibration and possibly make things worse."

Which is an unfortunate ideological position to take against any new entrants to the colour card market. I guess that's why you can't just define your own?

\$\endgroup\$
1
\$\begingroup\$

The main value of ColorChecker and similar cards is that the actual, "physical" colour of each cell is known. (Colour is not an objective physical phenomenon, but there are models to express it in physical units, or by matching with known physical samples).

Hence, the measurements in "CMYK" or "RGB" as such are meaningless. It's like if you measured the length of something and found it was 3. 3 what? You need a reference to a known standard, a metre for example. Likewise, RGB is a dimensionless value; "R" in it just denotes the dimension ("direction"). In contrast, standards like sRGB have defined relationship to physical light properties and can be translated to other standards. Numeric values in one standard (say sRGB) will not match the values in another, even in the same space (say, AdobeRGB), for the same colour. Furthermore, the same "colour" will have different RGB values depending on brightness (which is considered an independent dimension). With CMYK, things are even more complicated because the resulting colour will also significantly depend on the substrate (paper).

Manufacturers of the calibrated cards like ColorChecker have instructions (usually in the software) similar to "this cell, under a given lighting conditions, have such-and-such chromaticity. When displayed in sRGB, it should have such-and-such relationship between R/G/B."

So, to have a correct result, you (or your software) need to know the calibration standard of each link in the chain, from the card to your display.

Lacking any of that, your best bet is indeed to calibrate "by eye", card in hand illuminated by the light similar to the original photo. Once you found the adjustments, you can try to apply them to the next photo, but don't bet on them.

\$\endgroup\$
0
\$\begingroup\$

Clearly, the cards are used as a reference somewhere in the image which contains an important object which probably need to be represented with some accuracy. Without knowing what the objects are and what kind of requirements you have for colour accuracy there is nothing to suggest here.

If would be a good start if you could know the spectral reflectance of those colours and defined your requirements. CMYK values you mentioned mean literally nothing.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.