I have taken the pictures in the screenshot for a Timelapse, both are 4 seconds apart. Exposure time was 2s, ISO 400. I’ve adjusted WB manually in darktable, and apart from that I’ve only pushed the pictures by around 3 EV to show the difference.
If you look at the darker, bottom part, you will notice that the color various greatly. I cannot explain that. I’ve tried denoise and chromatic aberrations but nothing helps. Anybody knows what’s going on here? Camera is a Sony A6300.
Would you be willing to post the two raw files?
Absolutely!
These are part of a Timelapse series and actually most of the pictures look like the first one, then there’s a handful of pictures like the second one and so on.
I checked the embedded preview files, but they look exhibit the same difference in color.
For now I’m trying to fix it using darktable’s “color balance” module, but maybe there’s a better way. And I would like to understand what’s happening here, if it was a mistake on my side.
color_cast_01.arw (23.8 MB) color_cast_02.arw (23.8 MB)
I should mention that I used a Variable ND filter which I adjusted from time to time for the changing lighting conditions. I’ve done that before and I can’t remember that I had problems with that…
There’s no problem on the camera side. I opened both in my hack software, and only did processing up to the linear RGB, then scaled for visibility:
I’m only cursorily familiar with darktable, but I’d suggest you process one image, make a style from that processing, and then apply it to the others. There’s definitely something different in the processing of the two renditions in your first post…
That’s very interesting. Can you push it a bit more?
I can see it clearly in the trees and the hillsides in the bottom half of the images.
What got me thinking about the camera is that the embedded preview images (extracted with exiftool) show the same difference in color.
I just did a short sequence with the VND and I think I found the culprit… 
Haaaaaaa. ImageMagick to the rescue! 
I’ve subtracted the images:
convert color_cast_01.tif color_cast_02.tif -compose minus -composite diff.tif
Then I’ve subtracted the resulting image (diff.tif) from the second image:
convert diff.tif color_cast_02.tif -compose minus -composite output.tif
And presto. I love FOSS
4 Likes
Educate the rest of us please!
Sorry if that was unclear but the color cast was introduced by the variable ND filter. Apparently it’s a rather common problem for videographers and even high quality (i.e. expensive) VNDs exhibit this behavior.
Yes, I should have done that in the first place:
To the previous processing I added the same filmic curve (regular Duiker curve, not @anon41087856’s fancy one… 
), and now it is apparent, decided shift in color.
1 Like