Bug #12331

Black spots when using sharpen, color zone, and exposure

Added by Benoit Brummer almost 2 years ago. Updated almost 2 years ago.

Closed: invalid
Target version:
Start date:
Due date:
% Done:


Estimated time:
Affected Version:
git master branch
other GNU/Linux
hardware architecture:


When using color zone to bring down the lightness of some overexposed bits (in this case neon lights), the final image has some black spots if the sharpen module is activated and the exposure is turned up.

DSCF6202.RAF (32.2 MB) DSCF6202.RAF affected raw file Benoit Brummer, 09/19/2018 09:47 PM
(DSCF6202).jpg (10.5 MB) (DSCF6202).jpg example output Benoit Brummer, 09/19/2018 10:22 PM
DSCF6202.RAF.xmp (47.7 KB) DSCF6202.RAF.xmp previous xmp Benoit Brummer, 09/20/2018 01:39 AM
DSCF6202_01.RAF.xmp (69.4 KB) DSCF6202_01.RAF.xmp unbreak input profile technique Benoit Brummer, 09/20/2018 01:39 AM


#1 Updated by Benoit Brummer almost 2 years ago

example output attached

#2 Updated by Aurélien PIERRE almost 2 years ago

Reproduced here. You have to enable "clip gamut" [troncature du gamut] to sRGB in the input color profile module [profil de couleur d'entrée]. This is also dampened if you enable "5 passes" color smoothing [lissage des couleurs] in the demoisaicing module [dématriçage].

Finally, you shouldn't use a positive exposure compensation here (it's already overexposed by 1/3 EV), which will lead to clipped highlights and out-of-gamut colors (as you see).

The best practice here is to underexpose (with the exposure module) and fix the black point, then use the unbreak input profile module [correction du profil d'entrée] (thus, gamma correction) to bring back mid-tones in a safe way.

#3 Updated by Benoit Brummer almost 2 years ago

Thank you! That did fix it, and I tried the technique you mentioned and was able to recover more detail, although it took me more effort to get satisfying results (and I used some masks, I may be doing somethings wrong, the resulting jpg's are on and I uploaded both xmp's here in case you'd like to look at them, though I understand this isn't a help desk and I very much appreciate the help you've given me already!)

#4 Updated by Aurélien PIERRE almost 2 years ago

  • Status changed from New to Closed: invalid

I'm working on a simpler way to recover the dynamic range, with another mode for the unbreak input profile :

That should be easier and quicker in the future.

#5 Updated by Roman Lebedev almost 2 years ago

  • Target version set to 2.6.0

Also available in: Atom PDF

Go to top