I don’t think AP has been working on the color eq much since he introduced the notion…at least not from what I can see…also with greater understanding of filmic and highlight recovery I guess we will have to re-evaluate the gains to be had be introducing the sigmoid curve… Now that the saturation is not dependent on the desaturation curve and the boundaries defined by the latitude segment I think AP has proposed different defaults for using filmic starting with a 0 or lowest value possible latitude. Contrast of 1.0. And setting both compression modes to hard vs safe and finally starting with 0.7 EV as the default exposure… All the recent videos from AP and others have clarified many of the issues around v6 filmic and highlight recovery and artifacts so I guess maybe another round of comparison with the latest filmic used as intended and sigmoid might be worth reviewing…
That commit from Aurélien ‘rework scene-referred defaults’ seems to be cherry-picked into Darktable-master somewhere today by @Pascal_Obry.
It was the one commit I always cherry-picked into my own custom-builds from R-Darktable (after it got left behind a bit too much for me).
It’s now in by default, so that’s one less step . Now, the only different custom step I always do is rebase jandren/sigmoid-tone-mapping on darktable/master and check it out and build that.
So if this ever gets merged, darktable/master is the one-stop-shop for my toying again .
(Even @hannoschwalm work on sraw seems to be already merged, and this might be a game changer for my DNG workflow…)
I’ll try more images but noticed a weird interaction when LC was enabled…
Using a different image I seem to have isolated it to an interaction between 3 modules when active and some certain settings in those modules…makes me wonder if this is related to gamut mapping in some way… Sorry if this is noise due to my build and the incorporation of the PR before it is merged
Tried reproducing it here without any success. Could you check if it has the same problem with only the PR? It does seem similar to what was reported by jorismak earlier in the thread:
So far in testing, this module makes it quick and easy to get a good result - even from an image with significant tone issues. The controls are simple to understand and have an obvious purpose. My impression is it’s fantastic, very good work!
I did clear my build dir and redo it. I will see if I still have the issue…
Build seems mostly fine…
I only noted these two issue wrt reported or flagged errors…
– Could NOT find Imath (missing: Imath_DIR)
CMake Warning at src/CMakeLists.txt:346 (find_package):
Could not find a configuration file for package “OpenEXR” that is
compatible with requested version “3.0”.
The following configuration files were considered but not accepted:
I need to spend more time with it…but it seems to really soften highlight details. I might have just not hit the right combo of skew and contrast… I tried esp on a recent thread with the sunflower and overall it looks great with the preset supplied but on closer examination there was quite a bit of detail lost in the yellow petals… On. recent FB post @bnpndxtr has modified filmic v6 to remove the gamut checking part and he shows some nice results…
If highlights are reconstructed (if needed that is ) , filmic v6 new defaults seem to match very closely with sigmoid. Maybe i mean it the other way around , i never understood why people like this so much when the difference with filmic is so small.
Again , that is not to whine or to moan ! To each his/her own !
But sigmoid has two modes , one which removes the color from highlights and one which tries to keep it (in my experience, this is how I describe them :)).
And if i compare between filmic max-rgb (preserve color ) and luminanceY (turn to white but more contrast) they seem similar in effect .
In the end , in sigmoid i always had to mess with the two modes and with the skew and contrast slider to get a result i wanted .
In filmic i always flip between two modes and click ‘auto’ for white , and maybe a little nudge extra . Honestly , most of the time I’m messing with sigmoid just as much if not more compare to filmic. So i don’t see the use (for me).
So, is this me not liking sigmoid compared to others ? Or me just liking/understanding filmic more than others?
I find it weird that I’m going ‘mweh, ok’ where everyone else is wanting it badly . What am I missing?
(Also , what does removing the gamut checking yield ? I never understood it to be a problem ? I thought people had issues with how clipped highlights showed without reconstructing, and the often flat, low contrast look?)
I don’t want to hijack this thread, but briefly, there is a Yrg gamut check that didn’t seem to be related to this issue I was seeing. I left that alone. The one causing me trouble was gamut_check_RGB, and I disabled that essentially, as an experiment. I’m going to run with this for a while and see how it goes. I’ve been using this for several weeks on various challenging images and the results have always looked great. Later I may dig deeper into the works and see if there is something just awry with that function itself. I understand the goal, but the results look weird. I can see color that’s wonky with my eyes…
.