False Blue Color with color-calibration

I have noticed that many of my images are gaining a very strong blue cast in the sky area … This is not from the blue sky itself but appears to radiate from the tree branches.
The color is directly related to color-calibration and can only be reduced (not eliminated) by setting the adaptation to 'none (bypass).
There does not appear to be any other cause to this problem and so I have to use other methods to reduce the blue … very unsatisfactory.
I really would like a solution to this problem … otherwise color calibration (using normal settings) works just fine.
On this file I have taken-out most normal post processing so as to better show the base problem.


20210518_0017.RAF.xmp (10.4 KB)
20210518_0017.RAF (48.2 MB)
This file is licensed Creative Commons, By-Attribution, Share-Alike.

2 Likes

I have seen that before, too, and just given up on some edits because I couldn’t figure out what was happening or get any idea of how to “fix” it.

You have terrible chromatic aberration’s in that area…these are being magnified…also the standard camera profile for your camera may not be that good at using rec2020 improves it…

YOur image no edits with as shot WB

Taking it to camera reference makes it a bit better out of the gate

Using linear rec2020 as input

CC brings it back to as bad as it was with as shot…

Default filmic will really exaggerate the artifacts…

Remove the midtone boost and use no for chrom pres and its much better but in the end the artifact is in your raw file at the start…maybe a better input profile??

3 Likes

My solution at this stage is to desaturate the blue in color zones
Unfortunately since this is a Fujifilm format raw the chromatic aberration module only provides a very minimal effect.
Blue-Sky

Some of the tactics here might also help…

I just used a second color calibration module to desaturate the blue channel (locally).

20210518_0017.RAF.xmp (14.3 KB)

1 Like

So far the best solution is to use a color-zones preset that increases the lightness and cuts contrast. This kluge can at least be applied over all/most images.
What I wonder however is what is creating this problem that has appeared rather recently. Can there be a problem with the Fujifilm sensor layout that is not working well with the recent color modules?
Masking for each and every image would appear very unproductive despite the fact that it works.

For fun I pulled it up in ART and it still has those artifacts. Since there are options trying the Adobe DCP in ARZt and disabling the base table did make a big improvement

I am not really sure what is going on here … with the minimal of operations used and NO other color modules the blue only really becomes objectionable when the color calibration is used.
I am not sure if recent color calibration changes may be contributing to this problem.

If you look back at what I presented your image with as shot…so your out of camera settings was quite bad to begin with. Changing WB to reference d65 made the image look a bit better. Then adding CC just really took it back to what it looked like with as shot. Screenshots were provided. If you zoom and select original in the history stack…so little to no processing there is a ton of artifacts

1 Like

Dcraw and Libraw’s dcraw_emu both show the heavy blue borders at the top, and less noticable purple bordering at the bottom, and similar effects at the sides. Different demosaicing methods show the same problem. There seems to be no coloured-bordering at the centre of the image, though it is difficult to be sure in this image.

If this was a consistent blue border everywhere, I might suspect a programming error. But I think this is a fairly simple chromatic aberration caused by the lens. The camera has a 6x6 CFA, rather than the more common 2x2, which software may have problems with.

The parameter “-C 1 0.9995” roughly corrects the bordering, although this introduces a roughness in the blue sky, which suggests the “-C” option doesn’t work correctly for this CFA.

Another effect is that the sky is clipped in the blue channel. At the edges of the dark branches, sensels receive less light, which reduces values in all channels, bringing the pixels to a mid-tone, so the blue is more saturated.

Yes, I have done the same process and unless I limit the CC process to the area mainly around the sky the results are still unacceptable. Doing CC mainly on the sky provides great results but that is obviously false.
What I fail to understand is how … suddenly … I am getting this rather strange problem.
You suggest a better input profile … I am using standard-color-matrix as input and REC2020 as working. I have always used those settings without any problem

Have you looked at your image without cc and Legacy WB set to as shot…and noting the comments provided above by @snibgo …the color calibration does not even come in to play here at all and there are substantial artifacts… for me adding CC in the legacy workflow result was about the same as the image produce by the legacy as shot so maybe provide a couple of other images where you feel you have the issue and it can be inspected and evaluated to perhaps provide more information. CA is not unexpected in backlit situations… while I did not look closely using linear rec2020 as an input profile seemed to correct a lot of that issue with the blues…

1 Like

I suggest using this camera and lens to shoot a target such as an open newspaper, perpendicular to the lens axis, in focus. That should clearly show if CA is the main problem.

OK … I will do some tests tomorrow as you are suggesting and will post the results. Wierd that this problem is surfacing after using this same gear for several years.

I think this is his issue to a tee and I did suggest the new module as it is now in the master but perhaps @davidvj does not have it…

My take with dt 3.5.


20210518_0017.RAF.xmp (9.6 KB)

1 Like

Yes, that seem to be the same problem(s). The versions I use of dcraw and libraw are quite old, and upgrades may be better. There can be many issues with dark foliage against bright blue skies, hence the newspaper test to reduce the complications.

The newspaper test isn’t ideal, as it can’t test at infinity focus, but it should show whether CA is a problem, and provide a useful input for testing CA-correction software.

1 Like

My PhotoFlow + GIMP-Lab

2 Likes