I have experimented a little with the possibilities of reconstructing underexposed areas. Thereby I took some pictures of a sunset, where I tried to keep the burnt out area small, so the foreground is underexposed. I tried to process the image in Darktable, but unfortunately I did not get a really satisfying result. Here is the current (pretty screwed) state of my edits:
I have been using the new, linear RGB pipeline for some time now and am pretty satisfied with the workflow and the results, but with this image I am reaching my limits. First I tried to adjust the shadows and highlights with the tone equalizer, but I always destroy the whole image (Very flat colors and lots of artifacts). So I switched to working with a parameterized gradient filter and darkened the sky separately, but this creates some artifacts in the leaves. Out of curiosity I tried the whole thing in Lightroom, where I got the following result (without extensive editing):
Here I still have some details in the sun area and also less problems with the leaves in the upper part of the picture.
Do you have any suggestions on how I can darken the highlights better and raise the shadows without the annoying artifacts?
Raw file: 20200602-2051-6020376.orf (16.8 MB)
XMP: 20200602-2051-6020376_01.orf.xmp (10.4 KB)
Edit: By using RGB power norm in the tone equalizer I got it working and was able to fix the shadow recovery a bit, there are still some artifacts in the fine details of the leaves, but I think it is near the lr result now. I also changed filmic’s preserve chrominance to max RGB which brought the sun a little more in the direction of what lightrooms doing. I still loose details near the sun, but I think the overall look is better with the DT approach now. So right now the only real problem left would be the artifacts in the fine details of the leafs which are caused by the paramteric + gradient masked exposure change to darken the sky.
Updated XMP: 20200602-2051-6020376_02.orf.xmp (21.4 KB)
Edit 2: Added license:
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.