Difficult picture to adjust highlights

I have edited so far this picture to obtain a natural look, especially of the sun and sky. I appreciate any hint to improve the picture. You will find below the RAW and corresponding XMP file.


_DSF3698.RAF (26.9 MB)
_DSF3698.RAF.xmp (18.7 KB)

Yeah, hard to not posterize the sun…

From the linear rgb, I applied the librtprocess highlight recovery just to keep the blown part from magenta-izing. Then, the bog-standard Duiker filmic, not at all agressive. In rawproc:


Thanks for the raw!
That was challenging - not quite happy with the result, but I post it anyway.

_DSF3698.RAF.xmp (43.2 KB)

edit: Oh, some explanations as to what i’ve done:

  • segmention based HLR (with adjusted settings) gave good results for the clipped sun
  • to get rid of the yellowish sun colors, i have a high ‘preserve hue’ value in sigmoid (also high skew to get contrast in the highlights)
  • blur around the sun to blend colors more subtle
  • tone equalizer to recover more of the sun (and to bring light into shadow areas)
  • played a bit with some masked exposures to brighten up some point of interests
1 Like

Well I had even more problems as exporting consistently caused darktable to crash, so here is a screen grab of the preview.
Darktable 4.1.0+865~g70da6dcc8

I just used sigmoid with these settings.
Exposure +0.7
and boosted the saturation in the shadows and midtones by 20% in colour balance rgb, although this might not really be necessary.
Highlight reconstruction - inpaint opposed
Boosting colours in the highlights or playing with tone equaliser led to banding.
Here is the xmp, but it might not play nicely…
Adjust highlights_01.RAF.xmp (9.6 KB)



Here is my version, personally I prefer a more orange tone, but it is enough to decrease the preservation of the tone in Sigmoid.
In Sigmoid I adjust the contrast of the whole picture (mainly the middle tones) and then I recover the highlights and lowlights with the presets made with the help of the community of this forum.

_DSF3698_01.RAF.xmp (17.4 KB)
Greetings from Brussels,

Not easy at all … here is my try using filmic.

_DSF3698_02.RAF.xmp (20.1 KB)


hard case indeed. i found this image makes an interesting point for aurelien’s dtUCS.

vkdt pretty much default processing (other than see filmcurve params in screenshot), but compiled with colour reconstruction via

aurelien’s dtUCS:

dng-style hsl:

(ouch, i do need to make this an option or go and fix the remaining problems with the UCS)


how do you treat colour in the curve in rawproc?

Harshy. I apply the Duiker filmic equation uniformly to the three channels.

The input conditions deserve note. In my current workflow, tone curve is applied to the data in linear camera colorspace, with only black subtract, white balance, demosaic, and in this case highlight reconstruct to tame the blown sun.

_DSF3698.RAF.xmp (10.3 KB)

Bit surreal but hey…

Edit…maybe a bit more toned down…

_DSF3698.RAF.xmp (17.1 KB)


Thank you very much for several rendering/editing suggestions. I found the result achieved by @spidermonkey very realistic. He/she uses the sigmoid module (?) of the latest DT release. On my PC I use DT 4.0.1. released via the website that does not include sigmoid. I will examine the settings - xmp file is provided, thank you for that - as soon as a new DT version with sigmoid is available at the DT website.

Motivated by some good results I tried to improve my image. Here is a second try:

In the “filmic rgb” module “white relative exposure” slider is in this case very sensitive to change highlights around the sun. I could achieve this result almost using this slider. I also added some noise to highlights and # of iterations (here 1) of high-quality reconstruction plays also an important role.


The corresponding new XMP file is included.
_DSF3698_01.RAF.xmp (23.2 KB)


I like your resulting image (the first one; Aurelien’s dtUCS), too. But, I am not familiar with that DT module shown at right. :thinking:

I think there is still the possibility to make the sun interesting. I tried to adjust the transition to leave it as intact as possible and with a small ring/flare around it as you might see sometime…otherwise as some have chosen to do you let it blow out or you blur it and color it and end up with a blob in the sky… I didn’t spend as much time on the overall color and tone so my edits could have been better I am sure but I think to me the sun looks better blown out or with some sort of bright center if it can be managed rather than a diffuse blob of color to hide the fact that it is blown out… It will be nice to see even more edits and see how people handle the sun… the HLR sliders in filmic do offer some options which is nice. I tried this image in ON1 photo and it really had no answer but to colorize the blow out with a luminosity mask…

I might go back and try the guided LHLR and see if as AP presented it can be used with filmic to restore a gradient in the highlights… Thanks for a great image… makes me want to be at the beach

my version using dt 4.1.0+927~g5c7fab96e-dirty

_DSF3698.RAF.xmp (13.2 KB)


I am happy that I shared this photo for DT editing. It is incredible what you can do nowadays using a RAW file. 10 years ago the camera would just blow up the sun and the surrounding sky. DT is very powerful to handle such difficult images. I am still a beginner regarding DT I feel.

1 Like

_DSF3698.RAF.xmp (13.5 KB)

This is how far I was able to go on DT 4.0.1
But I really like the edits of @spidermonkey and @7osema

Out of curiosity - I did put the file in Lightroom Classic. But this is without any modification - simply import and export. And the items of interest to me are the blown out sunshine and the mountain below it.

Looking at the results from @spidermonkey - sigmoid based I am very interested to see the final product when it is released.

On the left - the history shows that nothing is done to the image.

Slight sidetrack from the subject. Adobe developed interesting AI based masks in Lightroom (subject selection / face detection etc.). While I do appreciate the masks in DT - If somebody has the knowledge how to do that - it would be quite a benefit (if ever - I expect it maybe a huge undertaking).

1 Like

My attempt

As far as I know, both sigmoid and inpaint highlight reconstruction will be available in 4.2 release.
I ended up having to do very little to the image to make it look good, so credit should go to the developers.
I tried one of the weekly windows builds of 4.1 using a different config folder and it all behaved nicely (except for exporting this file - perhaps because it is fuji).

Just to clarify, I think @hanatos 's screen shot is showing the new application, vkdt, not darktable… but that is about the limit of my knowledge - except that it is fast :smiley:

Just for context and support of how good the FOSS stuff can be…


Default opening

Then enabling their hue based highlight recovery…

They can seem to access the fuji profiles or their version…

Here is pro neg…

Just for context …