Tone equalizer vs tone curves

I was intrigued by one of Bruce Williams’ latest videos, where he compares the results obtained by the base curve, the RGB curve, and the tone equalizer. My summary of the video is:

  • The three tools produce similar results
  • The tone equalizer interface is less efficient
  • Therefore, he’ll use the TE because Aurelien says so, but he’s not convinced.

I understand the theoretical advantages of the tone equalizer (in one phrase: it’s not limited to operate in the [0,1] range). Maybe what is needed to go from pure theory to an appreciation of its practical benefits is an example where the old curve tools fail, and TE succeeds.

Is this a good idea? Does anyone have such an example?

The keyword is chromaticity. How do you ensure chromaticity constancy across luminance changes. Once you understand this, everything else follows. And the three tools don’t produce similar results.

It’s very hard to explain the benefit of solutions that solve problems you never saw.

1 Like

I think I know what you mean; I have seen chromaticity (and, I think, saturation) issues with the traditional curves.

That’s why I’m looking for examples – to train my eyes see the problems, and to help others (like Bruce) who remain unconvinced of the benefits of the TE.

In my experience, while TE can be fiddly with the masking, the results are much nicer, the mask blends nicely with surrounding tones, I don’t need a parametric mask at all.

5 Likes

I rarely even use the masking, either I use the TE to try to do a little random dodge and burn with the scroll wheel or I use one of the presets, tone curve, compress highlights or relight being the most common baselines and tweak the curve a bit if necessary in combination with sliding the exposure set point for the mask and the odd time alter opacity…if you don’t over think it then it is really pretty strait forward to use…I have played a bit with the norm as they do create quite a different distribution of tones in the mask depending on which one you choose…its a bit like the color preservation norms with filmic and how they alter things

1 Like

Agreed, the results are much nicer, especially with one of those photos where you successfully create a great mask without any trouble. It’s like magic.
There are times, however, when I really struggle with the mask, especially on tricky photos with low contrast.

Does anyone ever use the Simple tab? I almost never do. The only time I want to use it is if I want to adjust one node without affecting any of the others, because if I’m not mistaken, that’s only possible on the simple tab. I’m no UI expert, but I feel as if the three tabs could be merged a bit more. So, having the ability to only affect one node in the advanced tab, plus also the ability to create more of a straight line, like traditional tone curves. Then adding in the masking controls (exposure/contrast) to the advanced tab to avoid having to switch back and forth. Or adding the histogram to the masking tab because I like to see that while adjusting my mask.
Just some random thoughts from a user’s perspective. As I said, I’m not a UI expert so maybe there are good reasons to have it all separate.

1 Like

I hover my mouse over the image while the advanced tab is visible and use the mouse wheel to increase/decrease. So nice.

2 Likes

I second that it would be nice to see the histogram without going back and forth when making changes…maybe we are really supposed to be looking at the mask not the histo of the mask…

tone equalizer is currently the tool I have absolutely no opinion on, because the main part does not make any sense at all - at least to me. That scale of 0EV to -8EV in an unbound dynamic space with cameras that exceed those limits by default already but the files while in dt are mostly only reacting to 2 or 3 of those sliders isn’t even confusing - it is just heaps and heaps of unpredictable for me at my current location in the journey through the dt pipeline.
Maybe later.

meanwhile a few clicks in color balance rgb and my pictures behave like well trained dogs.

:upside_down_face:

Thanks for all the feedback! I’d like to reiterate my original request for help though: show Bruce actual examples where TE avoids problems that the curves introduce. Words won’t convince him, he ran his tests and he can’t see the difference.

I’ve been trying to come up with examples myself, but this week’s been a bit busy.

Except sometimes the mask looks ok and the grey bar under “mask post-processing” looks good, like this:

image

so you switch to the Advanced tab and it looks like this:

image

So you realize that it can’t be ok and go back for some more tweaks. I admit, I sometimes find it challenging to tell how good the mask is from the mask alone. I know I’m looking for smooth blurriness in tonal regions but preservation of the edges, but I do find the histogram to be a safer judge of how good my mask is.

To me the whole video was sort of a miss because he mistakes TE for being a 1:1 replacement for the other tools. It is not. Filmic sets the upper and lower bounds, something he does in tone curve by dragging in the points, then tries to do with TE, and sort of does, but TE works a lot better with filmic enabled.

Also note that he had a relatively flat and desaturated scene, and didn’t at all bother to set a color picker point and see what actually happened with a measurement. He eyeballed it and said “well close enough” and sure, you could probably get away with any of the three with an image like he uses.

Note that o was hard pressed to give up the LAB tone curve as well. I know it, I’m comfortable with it, I understand the way it works. But Filmic + TE is easier.I have old edits where the file is flat that I needed to use two masked tone curves to get it right. Filmic and TE let me add enough contrast that multiple modules aren’t necessary.

1 Like

Mainly depends on how you set your exposure.

I use tone equalizer before (underneath) exposure to tame the highlights in extreme cases. 0ev is then always your highest white point, since almost nothing has been done to the data yet.

But normally I do what most others do : activate it (let it sit in the pipeline where it sits itself by default), go to the advanced tab, and then hover your mouse over parts of the image you want to make brighter or darker, and give little nudges with the scroll wheel.
I don’t care what the values are reading, I just hover and scroll without looking at the tool.

Almost always this is used to make a darker part a bit brighter if filmic crushes it too much.

If filmics output is too flat I can also use one of the tone curve presets inside TE to give some contrast. Although then I always change the 0ev part (in simple or advanced) to be back at 0, so to not change the highlights that I have set already.
I do this if the image is too flat but I don’t want to mess with local contrast too much (faces / portraits for example).

1 Like

And always the big benefit of the modern tools is that it doesn’t destroy all the highlight data that is there but is too bright to be visible.

I think a normal old curve tool will destroy everything over 1.0 before filmic has a chance to map it.

If you use them after (above) filmic it’s OK, but for me you might as well use TE.

I also like that it starts as a flat horizontal line instead of a diagonal line up. I know this is against the convention, but for making adjustments it makes more sense to me. Easier to see what is happening at every point (making it brighter or darker).

And apparently it’s better at preserving the correct chromatic representation when the brightness changes, but my adjustments have never been too strong to notice that.

I wrote a long description here, but realised that the best explanation is: darktable 3.0 : dodging and burning with the tone equalizer - YouTube

Anyway, I won’t delete it:
When you use tone equalizer in any of the preserve details modes, it’ll blur the image to create a mask; you then use the various controls under mask post-processing to get it (the brightness of the mask) to fit into that 8 EV range. The use of the mask guarantees that pixels in an area (unless separated by an edge) are treated the same way: the same adjustment, defined by the brightness of the mask covering them and the adjustment you set for that mask brightness level, will be applied, thus local contrast (the relative brightness of pixels) is maintained.

Suppose you want to compress the dynamic range (reduce global contrast) a bit. This could be done with an inverse S curve: you’d lower highlights and raise shadows. If you apply that curve without detail preservation (that is, on a pixel-by-pixel basis), a bright and a dark pixel next to each-other will end up with less of a difference → local contrast is lost. With preserve details, if the area containing those pixels is dark, both pixels will be brightened; if it’s bright, both will be darkened. That way, global contrast is reduced (brighter areas become less bright, darker areas become less dark), but local contrast (the bright pixel’s brightness compared to its dark neighbour) is maintaned.

6 Likes

I sometime let it go one way or the other and use 2 instances and focus one on highlights and one on darker tones…at least for a complicated image as then you target those areas in the one instance and not try to fix everything with one…

3 Likes

I have experienced this many times.
I tweak the sliders to make the grey bar under “mask post-processing” cover much of the shown interval but the histogram under the advanced tab turns afterwards out to have turned very flat like in your example.
Can anyone explain why this is happening and how to avoid this situation or systematically change the settings so that the histogram (and mask) becomes useful?

For myself, I depend on the Tone Equalizer just about always.

The big difference between the tone/base curve and the tone equalizer is that the former are tonal adjustments, while the latter is a spacial change.

The curves affect all pixels of a particular brightness the very same way. The equalizer segments the image into areas, then changes exposure smoothly in these areas.

Say you reduced highlights and lifted shadows. With the curve, this will reduce contrast in every part of the image. With the EQ, contrast within each region will remain unchanged, but the overall dynamic range is reduced. This looks more natural in most images.

The downside is, the tone EQ can produce subtle halos if pushed too far. Look for them in the little preview at the top left of the screen. They can be hard to spot in the full-screen view. (And the halos are much better-behaved than e.g. Lightroom’s Shadows and Highlights slider).

1 Like

…and the halos can be reduced by adjusting ‘smoothing diameter’ and ‘edges refinement’ in masking, at the expense of local contrast. It gives you the choice of having your cake or eating it…

1 Like