This is an image with high dynamic range from the sun on top of the waterfall. I realize that the shadows are vastly underexposed, but this was necessary to protect the highlights in the sky. The shadows at the bottom of the waterfall have a magenta cast to them. From my reading, this comes from the shadow clipping, with the green having double the amount of pixels than red and blue, this means that the blues and reds are boosted first, resulting in the purplish cast. My camera is an OM-D E-M5 ii.
I was wondering if there are ways to treat or prevent this magenta cast in Darktable. I have tried to use color zone to reduce saturation in the magenta range, which works, but introduces black spots in the highlights (some artifact?). I realize that I should have done an exposure bracket, but I didn’t carry a tripod.
If you are using filmic then there can be a few ways to tweak the results that you might see but quite simply use CBrgb…color balance… set the mask in the mask tab by using the white and grey contrast fulcrum by using the auto pickers then go to the 4-way tab you might want to lighten the shadows a bit there …it does a nice subtle job if you set the mask and then draw a box with the shadows hue picker in a magenta area… It will apply some green to negate the cast… you can increase or decrease the effect with the chroma slider… you could also try this in highlights even though its shadows I find that the highlights can contribute a lot to an image cast… heck just play with all three in this manner just to get a feel for it… A trick you can use the same process but add 180 degrees to the hue to acutally select and enhance that cast… Also if using denoise profiled there is a bias slider that can adjust the balance of purple/green in the shadows so this could maybe also help a little…
Magenta shadows is usually a black subtraction thing. I processed your raw with my hack raw processor, libraw delivers a black subtraction value of 253 and that seems to produce neutral blacks:
Darktable might be using a less-than-optimal black value…
with a bit of low contrast treatment + denoising. different but not magenta i thought. i noticed some chroma denoising? in the OP render too. possible that dt’s post-blackpoint subtraction+clamping denoising throws off the colour here?
(edit: the rawspeed black value for this iso is 254, so yes maybe that’s a difference too, but small…)
Great image!
I’ve seen those reddish shadows before - I have a similar camera to yours. But I’ve never had an image where I had to fix them.
I first tried to use the 4 ways tab in color balance rgb, but it wasn’t really working out.
So I tried fiddling with the black levels, as has been suggested above, in the ‘raw white/black point’ module. And increasing the black level 0 by one, and reducing black level 1 by one made a massive difference. I learn something every day.
My edit uses sigmoid in a recent build of 4.1
Thanks, everyone, for some very nice edits of a difficult (to me) image. I never thought there would be so much detail in the sky that can be restored, while still being able to manage the color cast in the deep shadows. Now, I need to check out all of those .xmps and see how it was done. In my first retrial, I saw that the raw blackpoint made a huge difference, even when making minute changes. Something I have to read up on to better understand.
