3.0 How to get good results automatically?

Yes, I have been, but it seems to change the character of the image in too many other ways, not just blur reduction. I don’t always like that effect. Yes, the default LC doesn’t change the ‘look’ too much, but it barely touches the appearance of blur either. So I turn it up a bit, and … too may other aspects of the ’look’ start changing.

Maybe there is a way to use LC that clears up blur without giving a ‘cranked-up LC’ look?

Contrast equalizer? You can control the what frequencies you want to sharpen.
You can control the luma or chroma.

1 Like

That all depends on what sort of blur you want to get rid of (haze, lens, demosaicing, …). For lens blur, there is no deconvolution module for now.

Also, avoid pixel-peeping too much. Exporting a 24 Mpx raw at 4K resolution means you compress 8 input pixels into one output pixel, so evaluating the sharpness of the picture zoomed-in at 100% doesn’t make much sense.

1 Like

OK, thanks.

Contrast equaliser was suggested, and I am looking at it now. It is pretty complex — I will have to study it — but its ‘sharpen’ preset looks promising.

The blur I am keen to address is from demosaicing.



Let us know how you go with it.

The contrast equalizer is quite powerful. I myself use it a lot for chroma denoising.

I too was looking for a replacement to sharpen and I was also redirected to using contrast equalizer (CE). I am now using routinely softer versions of the presets “sharpen” and “clarity” as straight replacement to sharpen and local contrast modules. By softer version I mean that I reduce their opacity to 50/60%. You may want to have a look at those presets.

The sharpen preset in CE with reduced opacity does very little zoomed out and you only see the effects at 100%, which is what you’re after I think.

@aadm @T_N_Args You may know but many don’t that the last dark bar sets the limit of the finest detail in a particular image that we can see and it changes with the zoom level that you are at. So any curve changes above that actually have no effect (see around 15 min in the video). Once I discovered this I was able to lift the luma curve at the edge of this bar to enhance fine detail and get some really fine sharpness. As you mention I think the presets are a bit too harsh so if I do a quick application of one I usually knock it back…also I have found bumping up the middle detail can give a nice effect as well maybe as high as the first horizontal scale mark with a lift at 2 of the adjacent markers… This is an old video but a good one for the equalizer… https://www.youtube.com/watch?v=zzVXK4eAM5E


@aadm @T_N_Args Try to use bilateral verson of LC…often I like it better …seems sharper and doesn’t seem to brighten the highlights in the detail as does the default…in fact bilateral with a really low coarseness…and contrast produces some nice find detail enhancement…

Yes. I was using Local Laplacian LC because of M Pierre’s recommendation. But I do notice it very quickly applies an ugly ‘dark halo’ effect.

And I have noticed that Bilateral Grid LC does not do this — at all.

done - german is here: https://www.bilddateien.de/blog/2020-01-26-darktable-3-workflow-lab-oder-rgb.html

@aurelienpierre: Two things are unclear though:

Les modules qui travaillent en RGB linéaire et réalisent des opérations non-linéaires mais respectueuses de la chrominance (à condition d’activer le mode ad-hoc ) sont :

  1. courbes RGB ,
  2. niveaux RGB .

What does ad-hoc refer to?

Balance couleur :

  • CDL : RGB_sortie = (slope × RGB_entrée + offset)^power

What does CDL stand for?

The english version on hackmd.io is also unclear about these.
Thanks - great article, great background infos.


Ad-hoc is latin for “designed specifically for that purpose”. RGB levels and curves can preserve the chroma provided you enabled their “chroma preservation” mode.

Colour decision list, aka slope/offset/power mode in color balance.


Thanks for quick explanation. I added that als footnotes to the translation.

It just occurred to me that you can do the following:

Turn off auto-apply basecurve and auto-apply sharpen in the presets
Create module presets auto applying filmic, tone equalizer etc.

Works fine………:grinning:!

1 Like

Something else I’ve only just discovered that will significantly streamline my workflow…

It’s possible to create presets that are auto-applied to images but start off disabled. This means that you can override the default module settings but still enable those modules one-by-one. That way you can see the effect of each module in turn and tweak them at the right point in your workflow (without having to manually select a preset each time you move to a new module).

Just set set up the module as you like, disable it, then create your preset.

I knew you could do that with styles - didn’t know you could do it with presets.

Edit: Having used this on a few edits now, be careful because if you compress history stack and go back to an earlier stage, or reset the parameters, you’ll be back to the usual darktable defaults. I think it would still be nice (perhaps I’ll raise a feature request) to be able to change the actual module defaults so that they survive a reset and don’t appear in the history stack until switched on.


So… I see Rawtherapee 5.8 has Capture Sharpening, link. This is the sharpening that I have been looking for earlier in this thread, link.

Can DT do this, or anything roughly equivalent?


1 Like

That’s work in progress : Got an image problem ? Go see the image doctor.

What if you moved them after filmic, which it is said outputs between 0 and 1?

Because after filmic, we are display-referred. Not scene-referred. The tone equalizer is designed to be scene-referred, and has an internal way of linearly scaling things back between 0 and 1 without applying any kind of non-linear transform. The filmic scaling is highly non-linear by design.

I was thinking of rgb levels and rgb curves since they can’t control values outside of 0-1, and filmic’s output is between 0-1. Having read your article, Darktable 3:RGB or Lab? Which Modules? Help!, I’ve stopped using those two, but reading this post, thought maybe there was a way to put them to work.

Thank you, BTW, for that article because I was lost before reading it.