Hi @micha
I made Lab adjustment version of this script. It’s still a prototype.
1 Like
@yasuo
that’s fantastic!!!
Here only WB:
And now with your new “Lab Correction with Relative Color Area Mask”:
It has never been so easy to achieve the desired L a and b values.
My goal would have been something like a:17, b:20, but I couldn’t achieve that because the slider for a* doesn’t go far enough with -0.200.
Do you think it could be made even stronger?
But the wonderful thing is that it’s so incredibly easy to change the a and b values, and they have only very little influence on each other.
You say it’s still a prototype - I’ll test it eagerly.
Yasuo, a big thank you to you. You fit in very well with Alberto - you both do so much and so quickly for ART that I’m really happy to be able to use it.
edit:
And I have already reached my goal:
You just have to change “narrow <–> broad”, e.g. to 0.044. I still have to get to know all these sliders properly.
This tool works very well when it comes to fine details. A really color cast picture can not be saved with it, it must first be executed a WB.
PS: My target values for a and b are not binding. I just wanted to quickly see how the script works. It works wonderfully!!!
The desired values for skin tones are: a: 14-20 and b: 16-22.
Hello @yasuo
Could it be that in the CIELAB color space, the a channel is actually the green-red axis, not green-magenta.
In your scirbt i find this: a*(G—Mg):
And I see that the a-controller has little influence on the b-values. Very good.
But the b-control has a surprisingly strong influence on a-values.
I have made trial version2 of this script.
I suppose it may be side effect of the mask. Try it without the mask to see if it has the same effect.
The new relativecolorc_lab.ctl:
With Mask Target Color: Red > a has only minimal side effect to b. b has strong side effect to a.
With Mask Target Color: All > a has strong effect to b. b has very strong side effect to a.
With relativecolorc.ctl, the slider: Blue (Y<–>;B) has little influence on a, whereas the slider: Green (Mg<–>B) has slightly more influence on a.
Basically, you can also change the a and b in Lab quite well with this relativecolorc.ctl, but you have to constantly readjust, because of the side effects.
The curves in Color > Lab Adjustments work very well, there is practically no side effect.
If you have anything special to test, please let me know, as I would like to contribute to the success of the project.
Thank you for reporting this problem. I suppose the reason for this problem is that I used the OK Lab color space instead of the CIE Lab color space. OK Lab is a refined version of CIE Lab; however, its color axes are similar yet different from CIE Lab. Therefore, when you move the a and b sliders, it seems that a* and b* affect each other.
I use the OK Lab color space in my scripts because Alberto provides an RGB-to-OK Lab transformation function in the ArtLib library, which makes it easy to use.
If this doesn’t cause any inconvenience, I’ll leave the script as is.
The relativecolorc_lab.ctl is good, with it you can change the a and b values better than with all other scripts.
The mutual influence is not 0, but comparatively very small.
Of course you can leave the script as it is, I couldn’t find any errors.
Thank you
Hello @yasuo,
the new relativecolorc_lab.ctl works flawlessly, I’m really excited.
What I still have to learn is to find the right mask settings for portraits.
There are two completely different approaches:
I realize that you can make a much more accurate mask with this, with the mask view you can easily achieve that only the skin tones are captured:
But when color correcting the L a and b values, it sometimes doesn’t capture the face so well.
This mask method is completely different:
You can change the color of the skin and face very well, although the “Show target color area” looks much less precise.
And I noticed that you can use both masks in parallel, perhaps an interesting option.
Do you know your way around?
If you would like to use ART built-in Color similarity mask for relativecolorc_lab.ctl, set “Mask Target Color” to ALL and enable “Color similarity mask”. And also you can use my relative color mask and other ART built-in masks simultaneously.
1 Like