Issue with clouds in RT

For some reason I can’t get RT to render clouds as they’re supposed to. There’s always some kind of darker fringe/ around them and it seems to be ‘baked in’ the process. No matter what I try it’s always there, with or without processing modules enabled, more or less visible. I compared it with the OOC JPG, basic LR and DT results and it’s only occuring in RT. What am I missing here?
Sample taken with Olympus EM10 III.
clouds_RT
Comparison LR & OOC
clouds_LR
clouds_OOC

Hard to say what you are even doing or what is possible without access to your raw and pp3.

Thats chromatic aberration from the lens - have you got lens correction turned on? LR will do that automatically.

Actually it’s a manual lens without correction profile available so far. However darktable is also doing just fine without fringes even though it has no clue about it. Hence I’m wondering why RT is working differently

Are you auto applying one of the tone curves?? Is so, you could try one of the different flavours…it will change the image quite a bit and so might the chosen highlight recovery method in place…you could check that??

I have something similar, but more severe for one particular shot from Sony A7IV. It happens only if I use Sony DCP profiles from Adobe. If I use the AutoMatched curve then everything is fine. BTW the same issue happens with ART too.

There were 3 gallery images here with white clouds… I downloaded the jpg and raw… I wasn’t sure I could see it as shown in your example…but maybe there is more nuance to it…

I think it is indeed more an issue with that UWW lens and a certain kind of chromatic abberation. RT tries to solve it but generates this greyish fringe instead. I loaded a similar Voigtländer 10.5mm profile and it does seem to work much better.

It doesn’t seem logical that lens aberrations would show up in one app but not in another, but never mind.

As already said different input profiles could do it - I wonder if ‘none’ might provide a clue.

Only thing I’ve seen similar is from a non-RT method called Contrast-Limited Adaptive Histogram Expansion, voila:

SDIM0441-RT- conv-crop-CLAHE.

Original RAW converted in RT 5.8 default:
SDIM0441-RT- conv-crop.

Would be nice to access the OP’s raw and the .pp3 …

Looks very weird in this example - talk about halos! - but interesting.

Yes, I left it with default settings to bring out the effect. It can be toned down and altered in local contrast size (tiles).

https://en.wikipedia.org/wiki/Adaptive_histogram_equalization

Apparently, Siril has it:

https://siril.readthedocs.io/en/latest/processing/clahe.html

1 Like

Without the data this will be an endless guessing game.

1 Like

As others have said, its hard to tell without the raw file. But, if as you said, turning modules off doesn’t change it, it might be due to the demosaic algorithm chosen.
https://rawpedia.rawtherapee.com/Demosaicing

I recommend playing around with different options to see if it changes anything.

1 Like

Please supply a suitable RAW file to get meaningful answers instead of educated guesses.

1 Like