UKC

 No wonder I mess up WB in photo software. This makes my brain hurt. Thankfully a vectorscope with my video software helps fix fec-ups.

https://youtu.be/WADuXiMZxq4?si=deY55qLYLiZug2OH

Excellent explanation. Thanks for the link 🙂

Nice vid. I'm a poor photographer but know spectroscopy so using K as a starting point always struck me as odd (OK when using the sun, but odd with LED light sources as they don't follow the same wavelength distribution).

So we have light falling (with a spectral range) falling on an object that reflects these with different efficiencies.

We then have a detector that has coloured filters over the individual detecting pixels - with differing efficiencies and sensitivities (most Bayer masks are pretty awful). The glass also has an effect.

We then project this via another set of emitting pixels from a screen - or lay down inks with their own reflective spectra (print) which depend, in turn, on the light source they are illuminated with.

Our eyes detect these wavelengths (rather poorly) (budgies do much better - and your household pot plant better still) and our brain interprets them.

No wonder we struggle!

What would you suggest as an alternative? You're right in that a complete spectral distribution dataset for every source, filter and sensor (including individually calibrated eyeballs) would provide more accurate colour perception predictions, but that's just too cumbersome for everyday use by photographers. Most engineering relies on 'reasonable' approximations that are 'good enough' in their domain and I guess K (or better mired) are just that.

I suppose there's no alternative - it just seems to have come from a bygone age of sunlight or incandescent/arc lights

I can't think of any alternative to K. Anything except for a continuous spectra wouldn't make sense to me. A lot of photographs are taken under natural light anyhow. With artificial light, no amount of adjustment to the white balance is going to make it look natural since it never looked natural to begin with. High efficiency LED lights with ~200 lumens/watt have a particularly green and nasty spectra (in my opinion).

I usually start off by selecting a grey point from an object that I think should look grey (sometimes there isn't one though), and then fiddle with the WB until it looks 'right'. My monitor is reasonably well colour calibrated but sometimes I check rgb values, like is the sky that I think looks bluey grey actually that colour in the image. If something was really colour critical, I have a colour calibration card, but in practice I've only ever used that to calibrate for different camera sensors.

Very interesting. Thanks for sharing.