I attach an image with a similar situation. The image was processed using your style with just a small amount of exposure adjustment … nothing more. I then wished to develop cloud detail/contrast using the t-e and this worked well on the clouds but also modified the surrounding sky and landscape. I then tried to localize the t-e effect by using a mask. The (automatic) mask correctly selected the cloud area but that area then broke-up.
My intention of using the t-e module was to try to provide a similar outcome to the use of the shadows-highlights module. Without a doubt t-e does an excellent job as seen in this cloud example but the effect is not localized and for that reason I tried to use a simple parametric mask.
I used the 125 value on the understanding that this would allow for the effect to be localized but maybe this was my misunderstanding.
20200809_0034.RAF.xmp (7.1 KB) 20200809_0034.RAF (48.2 MB)
Two renderings, one more contrasty, the other less so. No tone equalizer.
Why do you think you need the tone EQ?
Here’s a version with tone EQ (I’ve dialled back filmic’s default saturation, as it looked way too much for me).20200809_0034.RAF.xmp (11.0 KB)
As far as first impressions go, the EQ version looks sharper.
That is probably fairly easy to answer. I have been using dt for many years (obviously without t-e) but have since its inception moved over, fully, to the filmic process and I had hoped that t-e would eliminate the rather ‘flawed’ shad/high module that I previously used.
t-e is a bit messy to set-up for each and every image but on a global adjustment basis it works well. My hope was that t-e could be used as a true dodging and burning tool where both small and large areas could be independently adjusted. What is not helpful is when adjustment is not local and changes are made that are not needed/required.
The very nature of the t-e very fine interface lead me to believe that I was simply making local changes, I was mistaken. When I tried to localize the effect, through masking, secondary problems surfaced … and that is where I am today.
Yes, I can use many other modules and processing directions … that is not the question or problem; I am simply hopeful that eventually I can use t-e as a simple burn/dodge tool in the same way that I have done in the darkroom for more years than I care to admit.
Thanks Aurélien for your thorough reply.
I read it several times, reopened the ToneEQ module and tried to look at it with a fresh mind.
I think I get it better, especially the masking part … which I quite missed last time.
Promised : I will not open the Zone system module anymore, will stick with Tone EQ and re-watch your various videos about it:)
Thanks again & keep up the fantastic development AND teaching work !
Yes, I also see that now. I think reducing the dynamic range meant I was more free to set the curve in filmic (I actually find the tone-EQ’d version overdone, but one can always dial back settings).
The exposure-invariant guided filter (EIGF) got merged in darktable master today. It should help with the masking issues described here.
The combination of the latest version of filmic-rgb and the new tone equalizer makes for an incredibly powerful development duo. I am truly very impressed.
Filmic-rgb now appears to fully automate the base development regardless of the dynamic range … all the tones simply fit! The latest tone equalizer now allows me to make any necessary final tone adjustments.
A couple of other comments on filmic. I do not use (or find helpful) the 2 “look+mapping” displays … possibly since I do not totally understand them. The “dynamic range mapping” I do find both useful and of interest but I would appreciate being able to somewhere on the system see the actual dynamic-range of the raw data if that is possible.
All in all … a tremendous advancement in digital image data processing … I thank you.
David
Testing the new Tone Equalizer.
First impressions on EIGF: powerful, must be used carefuly or it looks too HDR-ish !?
I feel the need to increase the smoothing diameter.
What is the advantage that EIGF is supposed to have on the old guided filter ?
Any advice on how to use it ?
Thanks a lot for your work.
Hehe, years of work starting to pay off 
The raw data is unfortunately lost when it comes into filmic, too many steps have happened and tampered with it, plus the RGB data is encoded between 0 and some value, which means it has an seemingly infinite dynamic range (because of that 0, but it’s only encoding).
The guided filter (from Kaiming He, 2008) has 4 drawbacks:
- it blurs more aggressively low lights than highlights, so it’s not exposure-invariant,
- it is more sensitive to horizontal and vertical edges, but less to diagonal, so it’s not rotation-invariant,
- the surface-blurring decrease in intensity as the blurring radius increases,
- halos tend to appear at small radii around edges.
The exposure-invariant GF solves all 4. Typical blurring radii are around 1-5%, while the original GF is better between 10-25%. As it turns out, it’s also tremendously faster (because of how it works, but also because @rawfiner found some optimizations in the original code).
When result is too HDR-ish, best thing to do is probably to increase feathering
Maybe could you give an example to see what you mean by HDR-ish?
Here are few examples of what I call HDR-ish look.
This is with the preset compress shadows/highlights - medium
This is with the preset compress shadows/highlights - strong
I’ve also made a comparison with the old guided filter, using the same mild sine curve and adjusting mask exposure and contrast so that the histogram span is more or less the same.
This is the old guided filter
and this is the new EIGF
It look like the EIGF compress DR more (look at the blacks) and therefore can more easily produce un-natural results (?)
Thanks, I better understand your point.
The guided filter has a flaw, which is sometimes an asset: it blurs more shadows than highlights. This is sometimes usefull to have better local contrast preservation in the shadows. What you call HDR-ish look here is just a loss of local contrast.
That’s why I made 2 sets of compress shadows highlights presets: 1 set with eigf (which keeps better contrast in highlights), and 1 with guided filter.
On my test, with the majority of images I used the presets strong and very strong presets on (which have details in shadows and not much in the highlights), I prefered much more the result with guided filter.
Yet, eigf should be much easier to use when you are not using a preset, and when not doing too strong dynamic range compression: the blurring won’t change as you change exposure, which is not the case for gf.
In general, when you loose too much of local contrast, decrease feathering, but beware of halos
Also, it’s up to you to use the curve such that the blacks are kept dark enough. Whatever the masking, you need to drive the tones to preserve a natural relationship between zones (e.g. sky brighter than the ground, dropped shadows darker than objects, etc.)
Stupid question: when tracking the master sources, how do I see the new tone equalizer options? Do I need to blow away my data.db?
As far as I know EIGF should be selected by default unless it’s an edit on which you’ve used tone equalizer before (in which case, reset module params).
I see "no, average guided filter, and guided filter " as the “preserve details” options. Is one of those EIGF?
No. When you say you’re tracking master, what version are you on?