Pink color in picture in DT but not in RT

[focuscharger]

I am using Darktable 3.1.0 and Rawtherapee 5.8
I noticed pink color in sky in a raw photo which shouldn’t be there.
I checked in RT just to make sure it is not my camera (Canon 7D) and there is no pink in RT.
What is happening here?

DT:

RT:

[heckflosse (Ingo Weyrich)]

Most likely the raw white point in DT is too large.

[focuscharger]

Thanks,

How do I fix it?
Is there some kind of global setting which I have changed accidentally?

[paperdigits (Mica)]

In DT, toggle Highlight reconstruction or reset the module parameters.

[focuscharger]

I just tried that but it didn’t fix it.

I thought I’ll go back to 3.0.2 but when I install 3.0.2 it fails to start with error

[init] database version of '~/.config/darktable/data.db' is too new for this build of darktable. aborting

Is there a way to downgrade from 3.1.0 to 3.0.2?
I want to see if it is bug in latest dev release

[heckflosse (Ingo Weyrich)]

@focuscharger

How do I fix it?
Is there some kind of global setting which I have changed accidentally?

Don’t know how to fix it in DT, but the symptoms (pink highlights) are quite clear to be caused by a too large white level. See here for reference:

Some input on "recovering skies"

From RawTherapee camconst.json: { // Quality A, ISO and aperture WL data by CharlyW at RawTherapee forums, missing samples safely guessed "make_model": "Canon EOS 7D", "dcraw_matrix": [ 5962,-171,-732,-4189,12307,2099,-911,1981,6304 ], // Colin Walker //"dcraw_matrix": [ 6844,-996,-856,-3876,11761,2396,-593,1772,6198 ], // dcraw "ranges": { "white": [ { "iso": [ 100, 125 ], "levels": 13480 }, // typical 13584 { "is…

[focuscharger]

Thank you.

I’ll go through that thread.

[paperdigits (Mica)]

You can’t downgrade DT database or XML files. There should be a backup from when you upgraded, though.

[focuscharger]

I moved whole folder including backup to new location and installed 3.0.2

Same issue but after reading linked thread I can now set clipping threshold to .80 and method reconstruct in LCh in Highlight reconstruction module and color goes away.

I saved that as favorite and I’ll test it by setting it to auto apply when camera is 7D.

Thanks for help

[paperdigits (Mica)]

You should take care to not clip your highlights, then you wouldn’t have to deal with it in the first place

[focuscharger]

True but I didn’t face this problem before.
I used to use RT mostly but after watching @s7habo and @Bruce_Williams 's videos I decided to switch to Darktable completely and discovered above issue.

I have enabled highlight alert on 7D to keep an eye when reviewing photo after clicking and also dialling down exposure when shooting in bright day.

Thanks, I learned something new today.

[rvietor]

There’s a module “Raw black/white point” that’s automatically enabled for raw files.
You can’t disable it, but you can change the settings, in this case, the white point.

[focuscharger]

Thanks,

I think that is better option than trying to fix with highlight reconstruction

[eobet]

@rvietor

There’s a module “Raw black/white point” that’s automatically enabled for raw files.
You can’t disable it, but you can change the settings, in this case, the white point.

Thanks for the tip! I find that dragging this slider is quite slow and it requires very small movements. This will become tedious to do with many images. I wish there just was a button which “automatically adjusted the slider to remove all pink”… or better yet, that the software did this automatically (like other software seems to do).

[rvietor]

@eobet

This will become tedious to do with many images. I wish there just was a button which “automatically adjusted the slider to remove all pink”… or better yet, that the software did this automatically (like other software seems to do).

That would be an abuse of that particular module. Normally, the setting for a particular camera is fixed (or only dependant on in-camera ISO setting). Ideally, you shouldn’t need to touch it, but for some camera’s, dt can’t find the proper EXIF items. In that case only, you have to adjust the values (and best make it an auto-applied preset, provided you set it up with the correct white point, it won’t hurt images where it’s not needed).

The pink/magenta colour is normal if you have over-exposed areas, where your sensor was saturated in one or more channels (*). To check for that, use the “checkerboard” icon under the image (in darkroom view), To correct it as much as possible, use the “highlight reconstruction” module.

(*: Simplified explanation: over-exposing usually happens for highlights, which are white. Clipping the signal there, and then applying the white balance amplifies the red and blue, but not the green. So white turns into red+blue==magenta, or pink)

[sgotti (Simone Gotti)]

@eobet You’re necroposting very old threads. Current darktable 4.2 has two new and greatly improved highlight reconstruction methods thanks to @hannoschwalm work.

If you have issues with highlights reconstruction with current dt version you should open a new thread explaining your issues.

[Peter]

Perhaps a better thread? Measure black and white levels?

[priort (Todd Prior)]

Can you confirm what is the value in the exif data and what is the value used by DT… if they are not a match then you can make a preset… I think some cameras may have diff numbers for different iso but for the most part checking this would be a first step as that value is key for the HLR module to do its thing without having to tweak it

If you post a raw file someone here can confirm it you are still having issues