Summer by the sea

Yeah. Nikon, or at least my current Nikons have full sized jpegs embedded. Never have a use for them, though.

A tricky one, thanks for sharing!

dt-dev, sigmoid and a lot of try and error :wink:

History in jpg.

4 Likes

Photoflow:


DSCF3938.pfi (55.0 KB)

2 Likes

2 Likes

@deanc Welcome to the forum! It looks like something is obscuring the scene. Random guesses: Was this taken through the glass of a window? Did the humidity moisten the glass of your lens?

@afre actually not. It was taken a from a viewing platform overlooking the southern archipelago of Helsinki on the island of Vallisaari. It was a bit of a strange day. Overcast with spells of blue sky.

Quite difficult to me. I tried two rectangular local spots in RT, but the transition zone came out too dark


DSCF3938_RT-3.jpg.out.pp3 (23.2 KB)

I see ISO was 800, so probably DR400 was used on this photo, which explains the underexposure by 2 EV.

2nd try without local spots


DSCF3938_RT-5.jpg.out.pp3 (15.0 KB)

Art 1.9.1-14, using quite some tricks, including para and area masks. I didn’t lighten the trees in the foreground too much, to keep the backlit situation more or less intact.

DSCF3938.1600px.jpg.out.arp (13.8 KB)


Darktable 3.5 (history embedded in the JPG), with Velvia LUT from https://blog.sowerby.me/wp-content/uploads/2018/02/Fujifilm-XTrans-III.zip
color calibration’s WB taken from the clouds. graduated density to bring the sky closer to the foreground. color balance rgb for saturation. local contrast and contrast equalizer for details.

3 Likes

Reading the OP I thought ‘It cannot be that hard right? Raise exposure till ‘the mids’ or ‘your subject’ looks OK. Yes, the sky goes through the roof, but that’s what filmic is for…’

Add a touch of ‘local contrast default settings’ to bring live back into it if it turns out too flat. Then I do my ‘clarity’ trick, by adding another local contrast instance in the bilateral mode, low contrast (1, 2 or maybe 3.) Crank the detail close to 300%. Add it on a parametric mask, with the preview mode on and drag the sliders, so you get (almost) the highlights and nothing else. A bit of feathering, and tweak the opacity, so the effect is not overdone anymore.

Not a true edit, but a good example of what ‘raise exposure and ignore clipping, enable filmic, tweak white slider to taste’ can do.

But maybe I’m not used to Fuji stuff, but there is a weird pattern to the noise that I can’t get rid of tweaking the demosaic settings:

image

And it’s also in the sky, which can’t be overexposed… and can’t be that noisy at iso800, right?
image

Bright regions in images have a surprisingly large amount of noise; in fact, the standard deviation of the brightness is larger the brighter you go. Normally it’s drowned out by the faster-increasing signal, naturally, and the SNR is correspondingly higher.

But the fact that skies are often primarily blue means that the red and green photosites don’t come into play very much, and the luminance noise is more visible.

And additionally, in situations where you’re trying to draw a lot of contrast out of a very flat sky, the noise can become markedly visible.

1 Like

But there also seems to be some sort of ‘squirmy lines’ pattern in the noise, at all levels.

Something I recognize from Amaze but then only in the green channel (and is related to the fact that some camera have green channeks that are even enough to be used together, and some don’t and they need to be kept separate).

But here it’s with all colors, that’s what surprises me. Normally I can okay with ‘linking green channels’ option in RT/DT to fix that, but here I couldn’t get it away. Is that an issue with the ‘markenstein’ demosaicing (I never heard of it before) or is that something you have to live with if you have a xtrans sensor?

X-Trans always has some degree of wonkiness whatever the algorithm.

Markenstein demosaic causes squiggly texture on noise. Lightroom is infamous for having squiggly texture on details.

In RawTherapee you can try dual demosaic to alleviate this, it’ll let you use a different algorithm that handles noise better in low-detail areas.

A second attempt:


DSCF3938_curves.pfi (26.3 KB)

My first attempt was really complicated - this one is surprisingly simple.

+3.6 stops of exposure compensation
this curve:
screenshot

and a few tweaks on the RGB curves to adjust the colour.

2 Likes

thanks for posting
darktable 3.4.1


DSCF3938_03.RAF.xmp (28.1 KB)

2 Likes


Developed with RawTherapee, than bracketed with Photomatix, as I use to do in such cases.
Histogram
The Raw is completely underexposed. Remember, you never can better a bad shot in post, you can just develop it. To get the right exposure you have to shoot in Plain or Flat or Neutral or something like that. If not, the camera measures in regard of the JPG and not in regard of the RAW.

DSCF3938.RAF.pp3 (11.9 KB)
here the corresponding RawTherapee pp3

Maybe you have set a problematic mode that causes a mismatch between RAW and JPG and for that reason you are not satisfied with your JPGs.

Could you elaborate on this? I have an X100F but my RAW files always look underexposed when opened in darktable, compared the JPG you see in the first post. I just thought that it is what it is, not that I had misconfigured something?

Fuji always ‘underexposes the raw’ when you have their DR modes on, and sometimes at certain iso settings anyway.

An ‘underexposed raw’ by 3 to 4 stops is not a problem at all. Overexposing even a bit of the sky is :).

That being said, half or a full EV more here would’ve helped a bit with the noise if you care about that. But any more and you would’ve gone over. I think the shot isn’t a problem shot at all (although in struggling with xtrans artifacts which I’m not used to).

Raising exposure by a lot is not a problem at all. Yes the highlights look blown out to pieces, but in the pipeline the data is still there. Just apply filmic afterwards and get the white level back under control. No need for any ‘recovery’ since nothing is clipped.