Desaturating the shadows

How would one approach selecting (varying) degrees of shadows and desaturating them?

grafik

6 Likes

Sorry, but it does not do the job.
The “neutral” diagonal curve could not be considered as a starting point.
How do you think to desaturate the shadows or to saturate the highlights?
The zero (“neutral”) curve must be horizontal. Please read also my question: I can't understand LC and CL tools

You can use the colour toning module in mode colour correction regions to do that
ccregions.pp3 (13.0 KB)

Well, it does the job here.

If you apply this CL curve:

CLcurve

you may see that we have tweaked the darkest pixels in the image (look at the lower, grayscale bar), and by dragging them down we have effectively lowered its saturation (look at the left bar).

E.g. This is a crop of an image (original, without CL curve applied):

and by applying the previous curve:

as you may see, the darkest parts of the fruits have been almost completely desaturated

(Don’t forget to turn on the tool)

2 Likes

No again.
The behavior you obtain in this way is just “something” about the relation between L and C, but it’s not “good”. At least is not what you obtain with other software that have such tool and it’s not what you obtain in GIMP by changing saturation with a layer mask based on luminosity.

Hard to me to explain.
HH or CC or even LL curve has a function y=F(x) and its neutral function y’=x it’s obviously a diagonal.
Now, which is the meaning of a diagonal where x and y are different and indipendent things?
This is well treated by LH and CH tools, made in the only way they can work.
Why not for LC and CL?
Which is the meaning of a (i.e.) 20,20 point where the y is C and the x is L?
A point with L=20 could have every values of C.
What I expect from a C by L tool is a C’=f(C) for each different value of L.

A curve C=f(L) does not have too much sense. Sure it has “a” behavior but not the one expected by a tool called C by L.

The evidence is that actually, in a 0-100 L scale, is not possible to decrease saturation for a 0 (do you prefer 1?) luminosity point, as it is not possible to increase saturation for a 100 (99?) L point.

I can’t explain better.

1 Like

That’s what the colour correction module does. Did you try the pp3 I attached? Currently the answer seems to be no again …

@bluc

Example: Enable color toning, set saturation to -100, set mask to L equalizer and draw the curve you want, like this (left is original, right is with desaturated shadows)

Then add a second correction to saturate the parts you want and set saturation to +100:
grafik

Draw the curve to define the regions you want to saturate, et voila:

4 Likes

It is I who conceived this type of heretical curve (about 6 years ago) , which in fact is not heretical.

This system works as well as a horizontal curve and makes it easier to use BĂ©ziers curves

It applies to any chromaticity for a given luminance, and the derivative function y’ = x has nothing to do with it

But as says Ingo, it’s probably not this transformation that you want to achieve

Jacques

4 Likes

@jdc Was going to say the same thing but it is better that you said it. :slight_smile:

Sorry, I did not noticed you attached a pp3.

Anyway thank you, but I did not want to know how to do what I need.
I only wanted to point out what it seems to me a bad (I mean no sense) bahavior of the LC/CL tool.

If you tell me it’s OK as it is, I’m fine. Thank you again.

@bluc The question from op was
How would one approach selecting (varying) degrees of shadows and desaturating them?

Then you asked: How do you think to desaturate the shadows or to saturate the highlights?

Please excuse I gave an answer :frowning:

@jdc I can't understand LC and CL tools
at the bottom of the picture, how SNSHDR work. Easy to understand what it do.
With the diagonal, you don’t have a good (I would say any) control at the edges, for shadow and hilights.

Anyway, I’m fine with your answer. You think it’s OK, I think not, but for what I paid I’m full satisfied :slight_smile:

I think I don’t follow you, but I don’t even care about formulae or graphs in a user interface. I just seek results: how do you remove color (desaturate) black? :flushed:

Again, how do you increase saturation of white, or an already saturated channel? I don’t mind what a user interface pretends to do. It simply can’t do that.

How do you increase 70 to a channel that already has a value of 254 (clipping at 255)?

How do you remove 125 from pure black (0,0,0)?

I think it’s not what somebody say here, but what could be done or not. It doesn’t matter what a user interface makes you to believe.

1 Like

About pure black and pure white I said 0-100 (or 1-99). 1 and 99 deserve their saturation level, but for all you wrote, I have to admit that you are fully right.
I was talking about LCH while I was thinking about HSV. My bad. :roll_eyes:

Thank you and sorry.
Also thanks and sorry to Ingo.

3 Likes

@jdc, hi Jacques, a question please.
Suppose you have the neutral 45 degree CL line and consider a point with x value = N.
Now you lower the curve in this area so that for x=N, y is now 0.9N, say.
What happens to the C values for all pixels with luminosity N? Are they -
a) reduced by a factor, that is, a multiplication
b) reduced by a fixed value, a subtraction, and bounded to zero, or
c) something else!
Thanks.

@RawConvert iirc it’s a)

As @jdc is an expert in Munsell colors, may I ask if that transformation is perceptually uniform? That is, the same apparent color, with lower saturation?

Applying a simple multiplication of channels doesn’t always lead to that results.

If I recall correctly, he factors in the shift with many LUTs.

@RawConvert
As says Ingo, it’s a)

@XavAL
the answer is in the use of “avoid color shift”

  1. if desabled, nothing is done !!
  2. if enabled
  • first, with relative colorimetry (personal algoritm whithout using LCMS), C and if necessary L are modified to be in the gamut
  • second, a Munsell correction is applied with 195 LUTf, essentialy where Lab is “false” : red-yellow, blue-pirple, and some greens. Munsell tables have the reputation of being the ones that best render color perception

jacques

1 Like