Custom camera D65 reference

Hi, in order to fix the incorrect WB temperature readings in CC for my camera (Sony A6400) I am trying to follow the procedure reported in the manual here.
The custom WB coefficients I get are R: 2.789 G: 1.000 B: 1.814 vs “camera reference” R: 2.835 G: 1.000 B: 1.499.

On the picture of my D65 calibrated display the CC reading is now CCT: 5024 K (daylight), vs CCT: 4489 K (invalid) I got before.

It seems all good.
What I don’t understand is whether this calibration is really needed to get correct colors in edits or rather it is just cosmetics, only to get meaningful temperature readings in K.

It’s not needed if you get satisfying results without it. But having better input results in less effort to correct bad input later in the pipe.

As described in the manual: if you want to use standard illuminations you need to fit to their prerequisites. If you do a custom color calibration or use a colorchecker then it doesn’t matter if your cameras coeffs aren’t far off…

The camera makes unknown assumptions to derive its numbers. The numbers you come up with are probably dependent on some asserted “neutral” patch from the image. I find the camera assertions to be a good starting point, and usually what I end up with in my final rendition. My hack software has no accommodation for color temperature; I think its more of a distraction in post-processing, where IMHO the attention should be on the RGB multipliers.

The Adobe DNG/DCP workflow uses a color temperature number to interpolate a scene-specific set of camera color primaries; I think RawTherapee supports it, but I haven’t tried it to assess its efficacy. Personally, it would seem to me that, if one is interested in getting colorimetric consistency for a particular shooting session, they should shoot a target shot in that scene and make a custom camera profile for the session.

I think I read this somewhere, can’t recall where, but that ‘invalid’ marking about the color temperature algorithm not providing a decent result in the particular range. So, I think your effort is probably more about soothing the algorithm than getting a proper result…

Direct quote from the darktable 4.0 manual:

When the CCT is followed by “(invalid)”, this means that the CCT figure is meaningless and wrong, because we are too far from either a daylight or a black body light spectrum. In this case, you are advised to use the custom illuminant. The chromatic adaptation will still perform as expected (see the note below), so the “(invalid)” tag only means that the current illuminant color is not accurately tied to the displayed CCT. This tag is nothing to be concerned about – it is merely there to tell you to stay away from the daylight and planckian illuminants because they will not behave as you might expect.

In other words: the coefficients provided don’t fit with either of the models behind the “colour temperature” definition. They still work to make sure a neutral patch is displayed as neutral.
This can happen for any CCT value, not just a particular range of values.

One example where you can get this: light reflected from a coloured surface, e.g. grass or leaves.

Note: even though I use a Sony camera, I rarely see that “(invalid)” tag on normal daylight images. If I get it, I either have a coloured (green) background taking up an important part of the image, or I have a “strange” lighting (reflected, or filtered).


And what that means in pictures I think if I had understood it from past readings is from this


THere is some threshold used by DT to invoke invalid when you are at a value calculated that lies beyond the threshold…


1 Like

Thank you for all the answers.
I went trough the video from AP where he explains the CC module and WB / camera calibration.
I did myself a color profile with a Datacolor Spyder Checkr 24 and I compared the color accuracy by using different methods on two reference images for my camera (from dpreview and imaging resource) which show a color target.
Here below are the results, which include an edit I did on Capture One (I still own an old license of it).

The conclusion is that indeed a custom D65 is not only cosmetics, but can improve color accuracy, even more if used together with a custom color profile.

where did you get the custom D65 for the cameras used to take the dpreview or imaging resource images? Just showing a color target also doesn’t give you any information over the illumination used to take the photo. So you might compare images taken under different conditions.
Creating a custom D65 reference for Wb module just makes sense for your camera - it’s not in general valid for other cameras of the same model.

Why? D65 coefficients are related to the primaries of the camera CFA, so I expect copies of the same model to have the same custom D65, within the normal production tolerances.

Sure but I used “as shot in camera” in CC in all tests and the CAT algorithm should compensate for the different illuminants, that’s its job isn’t it?

If you have any trace of metamerism in your color checker, you already have a problem with different illuminants (which is why the maker tries to limit metamerism as much as possible). The CAT algorithm cannot compensate for that.

In general, a calibration of that type is only considered valid for the illuminant for which it was done.

If each sensor behaves exactly the same you don‘t need a custom D65 value since that’s the assumption why predefined r/g/b coeffs are used in general

Your point across images and camera would hold and to what extent it would be hard to say without the physical testing…as for the custom values I believe that @MarcoNex is talking about shooting a white screen on a D65 calibrated monitor as demonstrated by AP in one of his videos to see if, and potentially improve things from what is provided by default from DT…but I might be wrong…

Exactly, and complementing that with a profile done with a color target shot in daylight.
He says that combo is reusable at least with good illuminants with high CRI. Plus I am trying to do this with all my cameras to secure uniformity in the starting point.
Of course with weird illuminants / mixed / artificial lights that would not do, but those are a nightmare anyway.

I usually look at such as providing a rich texture to a scene…

Kinda like the joke about the optimist and pessimist twins on their birthday; the optimist digging through the pile of manure: “There’s a pony in here somewhere!”