Up close and personal with the color balance module: An experiment.

Yeah. Because of that I, in general, most often refrain from giving feedback or am very polite about it.

Me on the receiving end of it: As long as the feedback is substantiated or a clear case of taste difference I really welcome it. Both Wocket’s and your reply made me have another critical look at my edit; There is room for improvement!

The funny thing is that I am the first one to point out over-saturated and unnatural looking images and now I produce something like this myself…

Lets just say that I suffered from a blinding case of enthusiasm and module inspired awe :wink:

1 Like

OK, I’ve had some advise and critique from forum and non-forum members. Some of it is a difference in taste and not something I can, or want to, change. I do realize that the critique I got was, at least for the most part, justified.

This made me revisit my edit and, with that critique in mind, I changed some parts. I’m convinced this is a better version and I hope I’m right about that:

dornbush.lighthouse.fujifilm.x.e2.raf.xmp (31.7 KB)

Thanks for the attention/critique/comments thus far!



I’ve found your first attempt a bit overdone and I noted the greenish tint as well.
The original, displayed here, left me with a feeling which can be summarized as “this needs something to pop”. I saw an unrealized potential. Still, it was a good composition.
Your second attempt is certainly better than both, to my eyes. I have no idea how it was done, though.
Could you share what exactly was your adjustments?
Did you use more than one instance of the module, for example?

1 Like

This was a darktable one-module experiment. Of course I’m not counting the 8 needed modules that are applied by default, although I did not adjust those in any way.

I only used the color balance module and nothing else to try and get a good end result (some might argue ‘decent’, but that might be a reflection of my current skill level).

I did, however, use 11 instances (10 in the first try).

All 11 instances have their own name/indicator to distinguish which part(s) they apply to. So if you want to you can dl the RAW, apply my xmp and have a detailed look.

The first change I made was getting rid of the sharpening attempt on the treetrunk. This was a disaster from the start :slight_smile: I used the same instance to slightly darken the trunk instead.

Second change I made was add an 11th instance to get rid of the greenish tint. I used an (inverted) mask from one of the earlier instances to apply this to everything but the sky.

I took three samples from dark, middle and light green, looked up the complimentary of those and used them as starting point. I also toned down the luminosity and added a tad of contrast.

Third change was the sky: I toned it down by adjusting the mask refinement and slightly changing the hue/saturation sliders.

Those were the only “big” changes.

I did revisit most of the other instances to tweak (moderate) them slightly, working towards a more natural look.

1 Like

Thank you, Jacques.
Your description of the processes is most welcome.

I will be working soon on one of my images using your “one module challenge” approach.
I hope I will understand more about this powerful and very efficient module.
On the other hand, I will probably continue using contrast equalizer for sharpening as I am very happy with it.

You’re welcome.

As I mentioned in my first post: This one-module approach is not something you want to limit yourself to in a normal workflow situation. It is a learning experience, or it was for me anyway.

The one thing I really missed: A way to sharpen in a precise/controlled way. In my normal workflow I would, like you, use the contrast equalizer module, with in my case assistance from a difference blend.


Sounds like very good idea. I will give it a try.

1 Like

Jacques, it’s an interesting experiment and I’ll have to try that myself. As you said, it’s been suggested that Color Balance could replace the Tone Curve as well as other modules. I’m skeptical of that approach because of the myriad controls - shadow, mid tones, and highlights along with contrast and fulcrum - necessary to achieve the effect.

I’m curious if you’ve looked at that in particular and what your thoughts are.

Nice work!

@aurelienpierre’s many references replacing module X by the color balance module in his Darktable 3:RGB or Lab? Which Modules? Help! article need to be seen in context of that article. Can the color balance module replace the tone curve on its own? No, I don’t think it can. But in a linear RGB workflow it can because it is part of this whole rgb exposure-filmic-tone equalizer-etc modules toolchain.

Most people don’t like change and rather stick to what they know and/or have experience with. I’m no different, but sceptical as I was I did jump on the linear RGB bandwagon (kept falling of a few times, too) and am now convinced that it is a way of working that I really like. I don’t understand all of it yet, but that path is part of the fun.

I almost wrote and am now convinced that it is the better way, but I think as long as it isn’t your bread-and-butter having fun editing is more important than this being the better/correct way.

I never went into this experiment looking for or expecting to find replacements for any specific module. Now that this exercise is finished I’m confident that this specific module lives up to the claims put upon it though.


I really don’t have a dog in the fight on Lab vs RGB, and I would rather use a tool where I understand what’s going on under the hood. At the same time, I do want to know that some alternate method is actually going to yield better results, especially if it comes at the cost of more time.

L*a*b* still has its applications. It is about knowing when to use which model and for what purpose. For most people, the transition to “RGB” is beneficial.


@Jade_NL Either I have not been paying proper attention or else there is something that I do not understand :frowning: Could you please show a workflow where the contrast equalizer module with blend mode difference would enhance an image?

I did not say with blend mode difference I said with […] assistance from a difference blend.

My preset turns on the blend mode difference. This will turn the image to black. I can now start sharpening and very slowly the coarse/fine parts start showing up through the black. When you think you nailed it: turn off the blend (click on the circle once). If you overdone it you can either go back or adjust using the effects slider.

I find this an easier way to determine the actual sharpness I want without being distracted by luminance, colours and all that.


U… er… what? can you do a screencast of how one were to do that? I can’t seem to follow… And I’d love to know how to use contrast equalizer…

1 Like

I’m definitely not going to make screencasts, maybe this will work:

We need a starting point so lets take this Play Raw entry image as a starting point.

Discard the history and open it in darkroom view, you should see something like this:

Expand, turn on the contrast equalizer module and make sure the luma tab is highlighted. This little sharpening how-to isn’t bothering with the other 2 tabs at all.

Select the difference blend by clicking on the circle in the mask toolbar at the bottom and select difference from the blend mode pull down menu. You end up with this:

The effect is easiest demonstrated when you use it over the full width (from coarse on the left to fine on the right). So put your mouse inside the square below the luma tab and use the scroll-wheel (towards you) to make it as big as possible.

Now go to the centre of the square and very slowly raise the bar. Look at the black screen and stop when you see the first sign of the cats right ear (viewer left). You see something like this:

Now turn off the blend by clicking on the circle.
Even though I used a real crude approach by using the ears and using the full width you can already see that this is much better then the first image:

Probably not that obvious this being a 1024x600 screenshot, but here’s a snapshot comparison:

The hardest part is deciding what needs to be sharpened and how much, but I really like the above approach a lot better then ogling it.

In this example I would probably focus on the finer parts (the hairs below the eye and next to the nose) by using the fine and some of the middle part of the spectrum. Something like this:

If the overall effect turns out to be to strong (or not strong enough):Adjust with the mix (effect) slider.

Hope this clears things up a bit.

EDIT: Replaced the 1240x600 “stamps” with 1920x1080 versions.


Did you do this with the hue color picker? I thought I read somewhere that when you make a selection with the hue picker, the complementary color is selected by the hue slider. However, by your wording, “looked up the complimentary,” it sounds like you used an external color wheel instead. If you did, may I ask how?

By “toned down the luminosity” do you mean adjusted the factor slider in the negative direction?

Nope, I did not. Your interpretation of my wording is correct and the CB picker does give you the complimentary colour. Why did I not use it in this case? You can only pick an area and not a single point.

I used the color picker module and RGBto website. As of late I started using RapidTables - Web Colors.

A good colour analyser/converter is a module I miss in darktable…

If by negative direction you mean attenuate/make less then: Correct.

Thanks for replying. I checked out the color sites you posted by using the CB color picker, selecting 175, 143, 79. RGB to identified it as #af8f4f, hue 40, and provided a complementary color of #4f6fb0, 79, 111, 176, hue 220. On RapidTables, I chose the Color Wheel Tool, https://www.rapidtables.com/web/color/color-wheel.html, and entered rgb(175, 143, 79). It identified it as AF8F4F, hue 40. So far, so good. The complementary color was #4F4FB0, rgb(79, 79, 176), hue 240, seemingly off by 20. Duplicating that page and entering rgb(79, 111, 176) resulted in #4F6FB0 and hue 220, the hue I would expect. RGB to appears to be the better choice for finding the complementary. Another site, https://www.sessions.edu/color-calculator/, provides the same complementary, #4f6fb0, but it doesn’t indicate the hue.

I was especially interested in your experiment since Aurélien Pierre recommends using CB instead of curves rgb. I tried to duplicate the result of using curves and couldn’t come close. Curves is so easy to manipulate. However, what you have done here provides a learning path, and I really appreciate it.

There will be, slight, differences across sites/applications. I advise you to read the color balance section of the darktable tutorial.

Here’s a quote that might be of interest to you:

During the HSL to RGB conversion, the HSL lightness is always assumed to be 50%, so the RGB parameters are always balanced to avoid lightness changes. However, during the RGB to HSL conversion, the HSL lightness is not corrected.

I’m assuming you have, but just to make sure: Also have a look at Boris Hajdukovic’s (he’s on pixl.us: @s7habo) color balance only Editing moment. My little experiment was basically inspired by that video and in time other stuff fell into place to make me actually try it myself.

I’m glad to find out that people find this helpful.

I have read it several times and the follow on:

As a consequence, editing in RGB, then in HSL, then again in RGB will not retain the original RGB parameters, but will normalize them so their HSL lightness is 50%. The difference is barely noticeable in most cases, especially using the modes that already correct the RGB parameters internally in XYZ luminance.

What it said to me was stick to one method or the other, and that RGBL mode was better. I switched to RGBL mode, but found it impossible to use because the slider varied from -0.5 to +0.5.

I worked through Boris’ video and saved Snips of each step, the image and the settings. Now I am going to study it again.

One thing is obvious, you are able to see changes in the image that escape me; although, I am 71 and my eyes are not as good as they used to be, and my monitor is 7 years old, although I profile it regularly. But, yeah, this has been tremendously informative. I’ve seen your work on other discussions, and greatly respect it. I always work through your xmps.