20191201_NIK1145.nef.xmp (7.5 KB)
Color zones
6 Likes
Well, three times I tried to export out of darktable and three times it stopped responding, so I give up. Instead, I’ll tell you my method. In channel mixer drop B slider in B channel to 0. This will give the image a yellow cast and fixing it by increasing R or G sliders ruins the skin tones. So leave them alone, and instead go to your parametric mask, G channel, choose input, and take it out of all but the darkest of shadows. In filmic you might have to drop mid tone saturation to 0.
How about this:
nasty.blues.nef.xmp (9.9 KB) darktable 3.2.1
I’m of the opinion that the blues should not be removed. There are blue lights that illuminate the scene so blue casts will be visible. That nasty looking (led) cast/glow should go!
3 Likes
I think it it is too much tuning for defective input profile.
To mitigate the blue cast, there is also the ACES trick that @anon41087856 used in some thread, but I am not able to find it.
@gaaned92 I think you are talking about these settings:
Red channel: 1.0 / -0.18 / +0.18
Green channel: -0.2 / 1.0 / +0.2
Blue channel: +0.05 / -0.05 / 1.0
Tried, but didn’t work for me in this case.
It is mentioned in this topic:
Yes that’s it. I think @anon41087856 said it has to be applied in REC2020 working space.
You can try that
R channel: 0.88, 0.06, 0.265
G channel: 0.033, 0.799, 0.362
B channel: 0.033, 0.033, 1
Changed the input profile to “linear prophoto RGB” and used color zones. Tried to avoid darkening of the blue areas which looks unnatural especially in her hair.
20191201_NIK1145.nef.xmp (7.5 KB)
Edit: And here is one, where the blue is only tamed (-70% mix in color zones)
2 Likes
Do you mean you think this is why it wouldn’t export? Darktable doesn’t like parametric masking in channel mixer?
Wow, thanks everyone for the excellent examples! I’m now going through a few of the xmp that have been posted above… I’m particularly keen on understanding color zones so that I can apply the same technique to other cases. I need something simple that I can learn & remember without having to do too much mental gymnastics (I’m obviously referring to the channel mixer trick that @gaaned92 and @Jade_NL are discussing above… I need to re-read the post by Aurelien “hardcore dt” first!)
Alright, so I went through a few of the xmps posted above and … I don’t understand them. At all!
First of all, I went through the settings posted by @sovereign; the process that removed the blue color cast is a single channel mixer with destination red and settings R+0.95 G-0.1 B0. Ok, if I replicate this on my original photo I get weird yellowish colors. I understand this may be due to the different white balance settings, but what I don’t see is how did he get to that point. There’s nothing intuitive here! Is there a reasoning behind? Did you fix the white balance first or what? How did you get to those parameters in the channel mixer?
Second issue, there was a discussion on input profiles above. I may be missing something but what is the problem with input profiles? The defaults are input=standard color matrix, working=linear rec 2020 rgb. Should they be changed? Why @gaaned92 made that comment about “too much tuning for defective input profile”? why defective? How defective? how can you tell it’s defective? Picture me with a big question mark and a dumb expression on my face.
Third, color zones as used by @yteaot. Finally, a single module that I can probably use it, no problems! … well turns out that I have problems also on this!!! I see how it’s been used (by dropping a big hole in the saturation panel right where the blues are) but again – I’m unable to draw that sharp rectangle .
Probably less ‘defective’, more about ‘deficient’…
3x3 matrix profiles are quite fine for the majority of images, but in the cases where there are significant numbers of colors out of the destination gamut, they tend to just pile up all those colors at the gamut edge. A camera profile that uses a LUT can be made to make those color mappings have more gradation than just ‘pile-up-at-the-fence’. That’s what’s going on with your blues…
Hi @aadm, you failed to notice I modified all the three color channels in the channel mixer module.
Just click on the destination and toggle between red, green and blue channels to see the values applied.
The rest of the edit is trivial and it revolves around exposure and tone equalizer.
@sovereign: thanks Matteo for pointing that out – I have indeed totally forgot about the other channels!
I still like your version the best so I’ll ask again: what is the process that you used to come up with those parameters in channel mixer? I can’t believe it’s trial and error, there must be thought process behind…
@ggbutcher Your explanation makes sense and I have seen that in fact just by changing the input profile to prophoto rgb the entire image changes quite a bit (I mean the colors); the histogram does change significantly.
I’m now using this profile for this particular image but then – what’s the rationale behind? That when photos are taken in difficult lighting conditions… then one must change the input profile? For what – one that is as wide as possible like prophoto? Is one supposed to change input profiles whenever… something happens? like shooting at night or in clubs?
I realize that I don’t even know what is the question that I should ask!
Finally, I’ll reply to the question I’ve asked @yteaot above about color zone module. With a bit more fiddling I have seen that to draw a square hole one must use the monotonic spline interpolation, alright! Then I have been able to replicate (more or less) what you did. I think this is the easiest tool to use even though I prefer the outcome using the channel mixer as I said earlier.
That’s why I’m working on the spectral profiles, ‘one profile to rule them all’. Getting camera measurements transformed down to the display gamuts is more suited to a non-continuous function captured in a mechanism like a LUT. If you dissect the Adobe camera profiles, that’s what they’re using.
Those values may come from Out of gamut colors, matrix, hsv - at least that’s the only post I found them (thanks for posting them here!). They do work remarkably well on this shot. I’ve added a bit of Lut3D and came up with this:
4 Likes
Ciao @aadm, what I did is a (very) creative color grading, and it does not make sense in every situation. In this case I replaced the blue channel with green (mostly) and red.
You can try the following to achieve in a few simple steps a very basic look that you can further refine:
- set the exposure to automatic
- activate the channel mixer and drag to zero the blue in the blue channel: the image will assume a yellowish cast because what you are left with are just the red and green channels.
- mix the green and red in the blue channel as you find pleasant; try for example 0,965 for green and -0,030 for red; aim for a nice overall color balance.
- activate local contrast with default parameters.
- apply denoise profiled and hot pixels removal.
Here you are the result:
As you see, the core steps are really few and simple.
The fine tuning might require some fiddling though …
To answer your original question, I didn’t do any math to set up the channel mixer.
I begun more or less in the way I described and then, touch after touch, I reached the final result.
To adjust the final color with the color balance was a sensible way to refine the result.
To adjust the white balance instead was probably not necessary.
Note: I am using the scene-referred processing style (filmic rgb activated by default).
3 Likes