Caveats for using Colorchecker with Color Calibration Module in darktable

@aurelienpierre or others… Apologizes if this has been mentioned or someone already asked…are there any precautions to pay attention to when using the colorchecker in CC?? So in the most basic setup I would guess people would have lens corrections, exposure and filmic enabled when they first open an image. Exposure is defaulted to 0.5EV. Should that be disabled before doing the correction?? Would other modules coming before color calibration be a problem?? Could this still be used with the legacy WB and CC in bypass and then just use it for color checker adjustment?? Basically just looking to confirm if there are any prerequisites or things to avoid before attempting to obtain the correct result when using the colorchecker calibration adjustments…Thx

Generically, for any target shot:

  1. No Glare!
  2. Underexpose rather than blowing any patch.
  3. Adjust white balance to a neutral/white patch.
  4. No Tone Curve. This is about the camera measurements, not what the target shot should look like.
  5. Straight-on shot, that is, lens centerline is centered and perpendicular to the target face, to aid in the success of automated alignment logic.

FWIW.

I assume you’ve looked in the dev version of the user manual?

2 Likes

Thanks Glen…taking the shot is not a problem…I am not sure it you are aware of this feature…its not a classic icc calibration but rather a tweak to the existing input profile based on tweaking channel mixer coefficients to create a better match to a color chart…so that is why I was asking assuming that modules coming before color calibration module may impact that…or perhaps some other thing I have not thought of

That appears to cover most of it but I was still trying to wrap my head around the fact that I assumed it would do the corrections based on input into the color calibration module. This would be fine given the setup explained by the manual but now if you add modules that come in the pipeline before color calibration would that not change the input to the module and therefore make the adjustments no longer correct…of course this must not be the case so I was just looking for clarification on the flow of the calculation…its likely obvious but I just have a moment of confusion on that process

I’ve still yet to get around to playing with it with the color chart I bought a couple of months ago but from the description in the manual I’d assume you want to keep everything at default values, with the workflow preferences set to scene-referred and modern.

The bit that might trip me up is

adjust the camera exposure such that the white patch has a brightness L of 94-96% in CIE Lab space or a luminance Y of 83-88% in CIE XYZ space. To be safe, you are advised to bracket your exposure between -1 and +1 EV in matrix metering mode, and pick the closest exposed picture during post-production.

which might have some dependency on other settings since presumably you have to use the global color picker to do this and afaik this measures the end of the pipeline.

Thanks for your comments…From making ICC files I am aware and the manual states about getting exposure for the patches as close to spec as you can so that was not problem and my assumption was that I would feed in that type of source image and therefore why I was suggesting to disable exposure rather than tweak it…from what I could see in AP’s video demonstration the code does some analysis to produce an array of channel mixer coefficients to further correct the matrix applied by the input profile to get a tighter match to color under the lighting conditions that the target was photographed…I am sure it is accounted for in the code in some way but being dumb and just assuming modules are processed in the defined pipeline order it just seemed to me that once you have this correction it would be modified if you added modules that impact color if they come before the color calibration…again I am likely being obtuse here I am sure this is just me being ill informed about the process.

Always room for improvement in the user manual if it’s not clear enough.

1 Like

For now I will assume its me not the manual…

I’m aware of the application; I just thought through what would be important to any colorchecker capture for any purpose, including the Color Calibration module. Second thought on the tone curve for this application; rather than that, exposure adjustment to put the data up against the top end might be in order, to put the patches in the mid-range without risking hue skew…

The best exposure and black compensation to use at calibration time can be found at the bottom of the profiling report. Basically, your colorchecker has a theoritical white patch value L = 96 or 94, so the closer it is from this value, before entering calibration, the more accurate it will be.

Exposure is a linear op (multiply), as the legacy WB and channel mixing are, so whatever coefficients are used should have no effect (apart from usual numerical sensitivities in the solver). However, the black compensation is an offset, therefore non-linear, so you need to nail it (and nothing guarantees that Rawspeed “raw black point” value is accurate on a colorimetric level, it’s merely the sensor reading baseline). But as soon as you changed the black offset, you need to adjust the exposure accordingly and the legacy WB gets also tied to the pack, so that whole set of settings becomes rigid.

Of course, at shooting time, you use whatever fits your hardware exposure.

Input profile and legacy WB are taken into account in the profiling, meaning you need to use the same values on the pictures where you later apply the profile. Honestly, if you keep the default scene-referred settings, you should be ok… That things as been designed consistently across the whole pipeline.

But, heh, you don’t need me here. The profiling report tells you the remaining delta E (calibration error) after calibration, so whatever minimizes this value is best. Just shoot several color checkers in different lightings, apply one profile done is some conditions to another shot in different conditions, and see how well it scales. You have everything it takes to do your own tests.

Thanks @aurelienpierre. I’ll try to remember to add something to the manual.

Thank you as always for your advice. I just reread the manual. I had noticed the exposure and black point in the report but had missed how to interpret that…