High contrast recovery

I was wondering how much better you guys can recover data from the shadows compared to my approach (while maintaining a decent overall look)!

This file is licensed Creative Commons, By-Attribution, Share-Alike.


DSCF1345.raf (18.7 MB)

3 Likes


DSCF1345.raf.xmp (31.7 KB)

4 Likes

Not much to change.


DSCF1345.raf.xmp (13.2 KB)

3 Likes

VERY FUNNY. I can see I know nothing of Hell xD (when it’s in the form of high contrast and underexposure and good taste, dealt with in RawTherapee).

So, I tried and I gave up, rather dissatisfied:


DSCF1345.jpg.out.pp3 (15.4 KB)

1 Like

ART. Noise in shadows is brutal:

DSCF1345.raf.arp (11.5 KB)

1 Like

This one is definitely a candidate for blowing out the sky. Otherwise the tree is always silhouetted. The image falls apart in the shadows because of the noise, and I didn’t really spend much time trying to clean that up. But this is what I got with AgX in the dev build of Darktable.

Darktable 5.1.0:

DSCF1345_01.raf.xmp (8.0 KB)

5 Likes

But your resolution is rather good, don’t you think?

I know fuji cameras have a lot of tolerance for underexposure but this one is pushing even a fuji past its limits, but here goes with DT5.1.
DSCF1345.raf.xmp (13.5 KB)

2 Likes

Solely looking at noise reduction, etc., yes, it’s not that bad. But from a colour management and overall interpretation viewpoint, I don’t know what to do with this image, and to be honest, I like the hues you’ve achieved in the sky.

And the saturation in your edit also suits the subject.

I hope someone used to this exercise will chime in — I tend to think there’s an approach better than others, for this kind of image, just from a signal processing perspective.

1 Like

I agree. I wish there was a mixture of the “for the sky” edit and the “for the tree” edit.

One more item: I notice it time and time again that the saturation is completely gone when having an extremely high dynamic range image. I understand that there is a - let´s call it proximity - effect that if you come to pure black and pure white there can´t be any color information. However, I thought this was a more of digital problem (aka not enough bit depth to give a suitable range of colors) mainly associated with jpeg. But since the raw files also do not contain a hole lot of color information at the borders I start to think about how such situations could be handled more suitably. I know there is HDR mode and that would probably solve this but if I just wanted more color not more dynamic range - is there a way to get that with a single shot?

My play in GIMP. An interesting exercise.

1 Like

I much prefer the silhouetted trees and the natural-looking sky, so the original one and the one from @Thomas_Do look the best to me.

2 Likes


DSCF1345.raf.xmp (29.6 KB)

3 Likes

Gorgeous. I will check how you did it.

I have no idea what you are doing. Mainly the tone curve LAB settings and your color calibration is very confusing - how did you come up with this approach?

This is a very basic edit, nothing new compared to the others.
agx, otherwise just denoise (profiled), diffuse or sharpen (2 instances, 2 presets), local contrast.
DSCF1345.raf.xmp (10.4 KB)

2 Likes

The LAB tone curve is mostly me experimenting to see how it affects color saturation, particularly in the Blue and Magenta spectrum, which I often adjust. In normally exposed images, this method can lead to oversaturation, so I usually reduce it using other modules like Color Balance RGB. I recommend turning it off to see how it impacts this specific image.

As for the color calibration, I use it to emphasize certain colors. By adjusting the individual R, G, and B input sliders along with their colorfulness, I shift the overall white balance toward a more neutral look. Of course other module such as Color RGB module or others shall work fine for this purpose too, but the way I see it the Color Calibration module can be used for more “Global” color tone trimming.

Base curve version for kicks…


DSCF1345.raf.xmp (15.7 KB)

2 Likes

Another variant: tone equalizer to tone down the highlights; one more local contrast to bring out more highlight details.


DSCF1345.raf.xmp (13.0 KB)

(You should really clean that sensor, @DanielLikesDT :-))

4 Likes