Note: further masking would be required in this case, as the pixel-wise exposure changes create harsh transitions and thus glitches:
Itâs possible. The mask contrast compensation was introduced because the guided filter blurs, thus reduces contrast. I have never seen any need to enhance the mask without guided filter, because the image usually has at least 8 EV of DR, so the whole (or most of) the controls are already used.
Depends. Of course, we make them as small as possible. Which, in any case, is better than just clipping blindly at the end to get the infamous rat-piss yellow sunset.
1 Like
By the way, that graph shows why Lab processing is bad. When you edit L at constant (a,b), that is at constant chroma, you basically shift colors vertically on that graph. While editing in RGB with exposure changes moves at constant saturation. Which gives much more plausible results. Even though it still doesnât account for Helmholtz-Kohlrauch effect, but thatâs another matter.
Sure, but here we have a case where someone wants to finely tune the exposure inside a small part of the DR; kind of âzoom inâ on the x-axis. Since the distance between the control points is 1 EV, stretching the relevant part of the DR over the available 8 ER is a viable solution. The mask exposure compensation already allows one to change the position (slide along the x-axis), the scaling would provide the âzoomâ.
Shall I open a feature request?
The code change above is obviously not enough: if someone with the current code tries several modes, first trying a guided filter, sets a contrast enhancement, then switches to one of the modes where the contrast adjustment is not used, a hard-coded value of 1.0 is applied by the formula. By naĂŻvely modifying the code, when reprocessing the image, weâd suddenly apply the contrast adjustment read from config, originally set for the guided filter, and thus the mask and the effect would change.
1 Like
Along the same lines, another useful tweak to the Tone EQ would be a local histogram, where the histogram displayed in the middle tab is only computed for the pixels included by the moduleâs mask. A situation I run into all the time is applying the tone eq with a drawn mask (parametric less often, but occasionally). The displayed histogram changes with the settings of the postprocessing mask in the third tab of the tone eq, but it still uses the global histogram even when using drawn masks, and isnât as helpful when trying to do this kind of local edit. Not sure how tricky that is to implement though or if thereâs some other workaround Iâm unaware of
1 Like
Or what about a color picker that lets you select an area as the method to set the mask in tone equalizer? That way in this example you could select an area of the sky and that would set the 8 control points based on the range within the selected area
2 Likes
There was considerable discussion about this âscalingâ in the tone eq not long ago I think. There was more than one argument why it might be difficultâŠI canât recall exactly what it was or which thread it wasâŠagain likely one you were part of ⊠
Yeah, I canât keep my mouth shut, can I? It was this one: [Suggestion] Simplify Tone Equalizer - #104 by pedrorrodriguez
But there the request was different, I think: instead of having to use the contrast and exposure controls to position the range of interest (the whole tonal range, or part of it) into the -8 EV ⊠0 EV range, somehow specify the ends of the range one is interested in, and the tone EQ would rescale everything to fit that that selection.
Here, my experiment was to enable the contrast adjustment for all modes, a smaller change.
1 Like
Opened a feature request: https://github.com/darktable-org/darktable/issues/10987
1 Like
3 Likes
Thank you, quite natural result to my taste, for the sky.
May you explain the key things you have made in DT to recover th clouds.
Is it with filmic or the recover module?
I have to analyze the xmp, but a brief explanation would be better to focus on the key concepts.
But why not look at the xmp first?
because yo may overlook which are the important steps oriented to get the work done in the sky, there is more processing in this photo that just the sky, and there are many âhiddenâ (not so frequently used) options in many of the tools that you can overlook.
Basically itâs an mix of âhighlights reconstructionâ module and âshadows and highlightsâ module. Iâm not using filmic in this case.Then color adjustments. Just check the XMP file.
I find the best way is to go to the active module tab as that shows the modules in pipeline order and then just work my way up the chain and observe the changesâŠnow with the little icon it easy to see which ones have used a mask so that is a hint to look a bit deeper on that one maybeâŠthats how I evaluate an edit of interestâŠ
Great to know, I could not get that clouds when using filmic, and highlights reconstruction did not work well many times.
Thank you I will study the xmp when I have access to my processing computer.
There is nothing hidden if you open the module and look at it. But it does take some effort.
1 Like
In filmic you can actually sometimes get away with boosting the midtone saturation all the way and then shift the latitude all the way to the highlights.,The effect will be too extreme but then you can now reduce the mid tone saturation until you get a nice level in the sky. It wonât always work but in this image I think it was not bad at creating a nice blue in the sky without too much impact on the image at least as I would have edited itâŠof course adding that locally with CB is likely a better approach but if you are using filmic it doesnât hurt to check it sometimesâŠ
DSCF0175.RAF.xmp (8.0 KB)
With 3.8.0. Disabled highlight reconstruction, used color balance rgb with the add basic colorfulness preset, plus a masked instance to desaturate the remaining magenta highlights. filmic with large latitude setting. Does not have as much detail in the clouds as some of the previously posted versions, but it looks reasonably ânaturalâ to me.
6 Likes