- 09/19/2016 : Final release !
Thanks for the testing, I’ll see if something can be done for these situations too since it’s related to the filter purpose. Be mindful if you add this to faves that it’s likely to get a few updates!
1 Like
@David_Tschumperle and @Iain,
In the final release of G’MIC 1.7.6 the “Saturation eq” filter “behaves” differently, also after refresh. In the pre release of G’MIC 1.7.6 the filter seems OK (also after refresh):
1.7.6pre:
1.7.6 final:
Also in releases before 1.7.6 it is OK, also after refresh.
Also there is a difference between 1.7.6 final and pre in the “Selective desaturation” filter. With filter in “reset mode” it looks exact the same as second image above.
also in “Graduated color accent” filter
Hello @iarga,
Thanks for reporting. I’ll check today and try to understand what happened.
I’ve done some recent changes/improvements in the color conversion routines, so that may be the cause of the difference.
1 Like
@iarga, I think I’ve found a critical bug in thr RGB->HSL function used by the CImg Library to perform the color conversion. Looks like I will have to release it asap, with a new version 1.7.7 !
Thanks again !
Yes, Thank you. I saw the use of RGB>HSL in the code of those filters. I do study a little bit now. 
Now I can change simple things, for example adding more color spaces to a filter. 
But from now on, I will follow your tutorials on the G’MIC website to get some basic knowledge. It will take much time. 
The bird has color again, thank you.
@garagecoder. Moving the Smooth slider for Detail mask (in Tone enhance 2) has a visible impact on the preview, but when I save a version with Smooth = 0 and Smooth = 20 they are practical identical. Is this intended behavior?
@paulmatth
Actually I’m not certain - it could be to do with smoothing not being scaled with image size. It’s possible that might not be fixable without changing to a different smooth method, I’ll check it out. Out of interest does it work for original tone enhance (under details)?
@garagecoder
Yes a bit better. There’s a visible difference between Smooth = 0 and Smooth = 20, the first is, ehh, less smooth and sharper.
I re-checked v2 and there are indeed differences to be seen, apparently I worked with a “wrong” photo that didn’t show those subtle differences.
But in both cases the preview shows more important changes.
I really like your filter, it adds a bit of magic to my photos!
@paulmatth
OK thanks, I’m afraid it’s a few hours before I can test fully! There are also pending changes - the filter is updated on git but not yet in G’MIC (need to wait for David to sync). The detail is now added after gamma to avoid issues depending on colour space, this could be in some way related.
I’ve just updated the filters on the G’MIC server, so it should be OK for you to refresh your filters.
I need to think about a way to automatize the updates in a clever way 
Okay, but situation unchanged: preview shows more differences than the result.
Ah, I think it has to do with the way the preview is rendered. When I zoom in to 100% and change the Smooth slider, I see exactly those tiny little details change.
Ergo, never trust a preview!
In that case the assumption about smoothing scale is almost certain - I should change the preview window to 1:1 default. The best idea is normally to scale smoothing with the image size but I don’t think that’s possible with -meancurvature_flow. Maybe I can add an alternative smoothing method, like I say though in a few hours