I added a curve (arbitrarily) L = f (H) to test the possibility of curve where the abscissa is hue. You can, if you tell me which, add any curve to the module “color and light”
For expanders, the principle is identical to those used in Wavelet
the enabled / disabled selection is stored
do not keep in memory open / closed
To change this (at least I believe it) it is necessary to make a change in the GUI of RT. It’s not in my skills!
To avoid crashing when the file path contains non-ASCII characters, I changed the location of the * .mip files.
Now by default, they are in “profiles” next to the “cache” folder.
You can choose the location, either “profiles” or as before next to the input file to be processed in:
Preferences / Image processing / Mip profiles
If you choose “profiles”, no crash problems due to NON ASCII characters, but if you have the same image file in different folders, there will be only one file “mip”
If you choose “Next input files”, the system crash with path with NON ASCII, but you can have one mip file, for each same file in different folders.
Normally at start-up, the system should not crash, but it will ignore the previous settings.
@dngimage.
I do not provide build, wait for someone to do it.
Is there a way that the Control Points can be noted better with a temporary ellipse or something similar, so the user can note or see the affected area(s) better? One work around I use is setting Lightness temporarily all the way down to -100 so I can visualize the affected area. Just a suggestion…
I push a new branch - “locallabgtk3”. This branch use GTK3 gui interface.
I make a change on this branch : now to enabled / disabled “curves” you must used a check button “Enable curves” instead of “lightness different of 0”
This function is for each control point.
@dngimage
What you are requesting, should probably be achievable, but requires GUI skills, which I do not have
When it will be possible again to “commit” a change, I would post something, whose performance surprised myself :)
For several months, I have been trying to perfect the shape detection algorithm. I think I have arrived at a very good result now, using the curves L = f (L) and C = f (C) - which are bugged in the previous version. While it is complex (at the code level), there are a few possible artifacts, but it works and there are two paliative artefacts that work well.
See these improvements soon
@dngimage
Thank you for testing
I think (I hope) I solved the bug for CBDL.
I just push a change to “locallabgtk3” with the new algorithm “shape detection”.
I call this algorithm “super” (instead of normal).
You must select for curves L=f(L) and C=f(C) ==> “Super”
and for slider “lightness” and “contrast” checkbutton “super”
In some cases (grey tones, flat areas…often with noise) these new algo can leads to artifacts:
use “global quality = enhanced + chroma denoise”
use the 2 new sliders “Threshold chroma” and “iterations”
It is slow on startup because it is slow on creating editors. That means when you are working in SETM (Single Editor Tab Mode) it’s slow only on startup but when you are working in METM (Mutliple Editor Tabs Mode) it is slower than gtk2 build each time you open an image (because that creates a new Editor Tab)