Before discussing the curves a* and b*, I’ll return to the ‘Lab’ module. I think I can speak about it (somewhat) seriously since I was (90%) responsible for developing this module starting in 2011.
The curves were added gradually; I don’t remember exactly when it took on this appearance, but around 2013 or 2014.
You’ll find essentially the same thing (except for the a* and b* curves) in Selective Editing > Color & Light (advanced), with the advantage of being able to define precise areas and use deltaE (It’s present even though it wasn’t broadcast as early as 2017.)
I’m going to start by clarifying Lab. Some say it should be thrown in the trash, that it’s crap, for a whole bunch of reasons - both true and false.
The only thing that is 100% true, but has nothing to do with Lab, is that even today, the output to our screens (which dates back to the 1980s) is still (by convention) using ICC profiles with 8-bit Lab… It’s a problem of standards and hardware compatibility, but has nothing (or very little) to do with Lab.
Lab is a transformation of XYZ (which no one disputes) by adding a gamma to Y. Simply put, Lab corresponds in Abstract profile to a choice where ‘Slope’ is 9.03 and gamma = 3.0. We work with 64-bit or 32-bit real numbers… There is no loss… Of course, the reference frames change, and this must be taken into account.
I tried it on images with a 25Ev Dynamic Range (I couldn’t find any with more), commonly the best cameras, pass 14Ev. There is no problem.
The major problem that emerged as early as 2011 is that of color shift when changing the saturation (an inaccurate term, but understandable). There is a strong shift of colors towards reds/oranges, blues/violets, and to a lesser extent towards greens. This can wait in extreme cases for a Hue change of 0.05 radians, which is huge.
I proposed a solution, which at the time caused some people to howl with indignation at its complexity. I created 195 tiny LUTfs (based on Munsell comparison) , loaded at RT startup (I thank Ingo for improving their handling in terms of ‘speed’).
Without you even realizing it, unless you disable it (it’s also found on some RT forks), as soon as you use Lab (including any Selective Editing), the correction with Munsell LUTs is implemented. The time spent is almost negligible. Admittedly it’s not ‘perfect’, there are still tiny discrepancies imperceptible to the eye.
One note: as soon as you use Cam16 or JzCzBz, it is deactivated because these modules compensate for Hue variations.
The other real problem, though it shouldn’t be a deal-breaker, is that if the user pushes the settings too far, nothing prevents the generation of imaginary colors. So what? In most cases, the conversion to output works correctly.
But these deviations are mere trifles compared to those introduced by exotic lighting (LEDs) or by manipulating the primaries (if you’re using Rec2020), when you adjust the ‘Attenuation’ (or lower the saturation), you enter the realm of imaginary colors. Nothing serious, but it’s good to be aware of. So, in summary, the potential imaginary colors of Lab are the same issue as in other cases.
I’m returning to the problem of using ‘a’ and ‘b’ . Personally, I don’t use them, nor do I use curves. I take very few personal photos. I focus on the most complex images to find, and I try to innovate by creating new algorithms.
There are several types of tools to do this (non-exhaustive list) :
-
Gamut compression (Leds, Sunset…)
-
of course curves Lab
-
Primaries & Illuminant in Selective Editing (with to see the CIEXxy diagram)
-
Primaries & Illuminant in Abstract profile - more complete with the possibility of working directly with the CIExy diagram (you can see that you are going outside the diagram, unless you are blind),or in ‘linear’ or in ‘polar’…The problem with primaries beyond imaginary colors is that they are not intuitive (at least for me). But I’m old…
-
These 2 ‘primary’ modules, because we work with a virtual (or internal, or abstract) ICC profile, allow us to easily adjust the illuminant, for example, going from D65 to D50… or something else), or the dominant color.
-
The new module that was recently merged is the one I called ‘Red Green Blue’ in Color Appearance & Lighting (CIECAM). Basically, seen from a distance in a tunnel, it does similar things (I’m not saying the same), like primaries or film simulations. In addition to the lesser-known capabilities of CIECAM (try it!), for each R, G, and B channel you can modify the hue rotation (h), the saturation (s) in the CIECAM sense, and the Brightness (which is only remotely related to Lightness) using a curve. As a reminder, when you modify the brightness curve, the apparent chromaticity changes according to a simple formula: C = saturation * saturation * Brightness. Of course, if there’s demand, I can improve it, but that will complicate the currently simple use. The risk of these imaginary colors is minimal because they are managed by the three processes (‘scene’, ‘image adjustments’, ‘viewing’) of a CAM (Color Appearance Model).
-
And finally, often overlooked, is the module that must be handled with care, “Raw Black Point,” which alone can solve difficult cases. However, its use is risky.
-
etc.
Jacques
