Desaturating perceptually for B/W looks?

Discussions - Using ACES Post (DI, Edit, Mastering)
1

Hi all,

Sorry it isn’t really about ACES but I was watching some videos about color spaces and came to this interesting talk. https://youtu.be/AS1OHMW873s?t=1797

He says that it’s usually best to create a black and white image in CIE-Lab because it gives the most perceptually consistent result w.r.t how we perceive given original colors’ brightness.

That made me want to learn and experiment upon which I first discovered that you get the worst results when doing this in log especially with very saturated(bright?) colors.

Pipe: RWG/Log3g10 → RGC → conversions → Resolve saturation slider → ACES rec.709 Out

Original

rec709 rgc_1.2.1
rec709 rgc_1.2.11920×1080 168 KB

RWG/Log3g10 desat

log3g10 desat_1.2.2
log3g10 desat_1.2.21920×1080 79.9 KB

ACEScct desat

cct desat_1.2.3
cct desat_1.2.31920×1080 67.6 KB

ACEScg desat

cg desat_1.2.4
cg desat_1.2.41920×1080 129 KB

RWG/Linear desat

rwg linear desat_1.2.5
rwg linear desat_1.2.51920×1080 126 KB

Clearly doing this in linear gives a more sensible result to my eye. Different color spaces still can give very different results.

Now back to the main question. Doing this in CIE-Lab… how would I go about this? I work in Resolve and I’m not sure how to apply the correct math to do this operation. The normal Saturation slider will result in something quite funky. I’m thinking it has to do with the fact that a, b are balances between colors instead of single vectors so to speak? So making them equal still returns colors?

lab desat_1.2.6
lab desat_1.2.61920×1080 142 KB

Would love to know some more about this.

1 Like
2

Turn off S curve option for contrast curve in project settings.
Make Color Space Transform sandwich going to and from Lab space.
Bypass R channel in the node inside the sandwich. Then set Pivot to 0.5 in this node.
Now contrast knob is saturation knob in Lab.
I’m talking about old contrast and pivot, not the ones from HDR tab.

3

That does indeed do the trick thanks! If you only intend to set contrast to 0 it doesn’t matter if it’s set to S-Curve or not right? anything lower than 1 is linear regardless? And if in non-managed your timeline space is set correctly you can just switch the node’s color space instead of creating a sandwich correct?

4

Yes, it’s linear in this case. Unchecking S curve only needed if you want to add saturation in Lab.

It should work. I didn’t try that with Lab, but it for sure works with HSV saturation for ‘density’.

1 Like