Project

General

Profile

Bug #10962

Strange blue colors when using presets in base curve module

Added by Martin Vala about 2 years ago. Updated about 1 year ago.

Status:
Confirmed
Priority:
Low
Assignee:
-
Category:
-
Target version:
-
Start date:
03/22/2016
Due date:
% Done:

10%

Affected Version:
2.0.2
System:
Fedora/RHEL
bitness:
64-bit
hardware architecture:
amd64/x86

Description

When i use preset in base curve i see strange blue color. Without preset i don't see such effect. Screenshot with preset(using_prest.png), without preset(without_preset.png). Source RAW file (20160321-190521-00.ARW) and JPEG directly from CAMERA (20160321-190521-00.JPG)

using_preset.png (2 MB) Martin Vala, 03/22/2016 09:41 PM

without_preset.png (2.09 MB) Martin Vala, 03/22/2016 09:41 PM

20160321-190521-00.JPG (4.47 MB) Martin Vala, 03/22/2016 09:46 PM

20160321-190521-00.ARW (19.8 MB) Martin Vala, 03/22/2016 09:46 PM

IMG_9524.JPG (5.16 MB) Ilya Kruchinin, 04/25/2016 02:45 AM

IMG_9524.CR2 (21.8 MB) Ilya Kruchinin, 04/25/2016 02:51 AM

lighttable.png (835 KB) Ilya Kruchinin, 04/25/2016 02:53 AM

darktable.png (938 KB) Ilya Kruchinin, 04/25/2016 02:53 AM

History

#1 Updated by Pascal de Bruijn about 2 years ago

  • Status changed from New to Confirmed
  • % Done changed from 0 to 10

This is more or less expected behavior.

The default camera brand specific basecurve tries to approximate camera tonality, but also affects color rendition (saturation), because it operates early in camera RGB, so camera's which produce fairly contrasty images by default, also tend to look more saturated in darktable by default. Depending on the type of image, that can be more or less flattering.

Feel free to switch to a different basecurve (there are other presets available) if the defaults results don't suit you. Or tweak to your own desire. Or disable it, and tweak tonality/saturation to your own liking using some of the other modules we have available.

#2 Updated by Martin Vala about 2 years ago

Dead Pascal,

I tried all presets and all of them look the same from saturation of blue color point of view. Can you give me hint, how to disable basecurve module by default? When i import images, it is there by default.

#3 Updated by Ilya Kruchinin almost 2 years ago

I have the same issue with Canon EOS 600D images.
The processing of blue colors is way off.
Selecting any base curve doesn't help.

This has been an issue for several releases (same issue on 1.6.8 and 2.0.3, but it dates back even further).
I tried disabling colord completely, which did NOT fix the issue.
Absolutely any image that I open is color-corrupted, and I am unable to fix it.
Had to stop using darktable as a result.

#4 Updated by Ilya Kruchinin almost 2 years ago

I believe the bug is due to incorrect handling of input color profile, or enforced color management.

Attached are off-camera JPEG, off-camera RAW, view of the file in lighttable (before going to darktable), and darktable view.

This cannot be corrected with a different base curve. This can be made less severe with a different input color profile, but the colors are still way off.

IMPORTANT: I should have probably opened a different bug report, as I don't believe the original name of the report reflects the true issue.

#5 Updated by Martin Vala almost 2 years ago

We can rename title, if needed.

#6 Updated by Ilya Kruchinin almost 2 years ago

Martin, let's first wait for some feedback from the devs.

Developers: Do you need any additional information? If necessary, I can record and upload a video with the bug.

#7 Updated by Michel Corps almost 2 years ago

Ilya Kruchinin wrote:

Martin, let's first wait for some feedback from the devs.

Do not expect an answer from Pascal on this subject (other the one he made already). A few months ago I reported the same "blue color problem" with some images from my Olympus OM-D EM1 and I got the same answer : "Feel free to switch to a different basecurve" which did not solve the problem.

So I avoid to shot blue subjects and this way I can enjoy this fantastic software ;-)
(I am joking, the problem is only about images with a particular blue and I really love DT)

#8 Updated by Pascal de Bruijn almost 2 years ago

The thing is that there are no easy answers here.

The some basecurves tend to exacerbate the issue, thus my suggestion to try others.

Another trick might help in some cases would be to enable gamut clipping in the input color profile module.

As far as color profiles are concerned, we essentially get these from Adobe (via DNG Converter), and there are no alternate places to get them from, as vendors do not disclose their color processing methodologies.

#9 Updated by Pascal de Bruijn almost 2 years ago

Also keep in mind that Darktable applies any active display profile by default (unlike many other programs, who incorrectly ignore it by default), thus if you have a broken display profile loaded it might affect color rendering negatively as well.

Keep in mind that many modern desktops enable a socalled EDID profile by default these days, which is sourced from the display hardware, sadly hardware manufacturers being hardware manufacturers this "profile" is regularly of questionable quality.

More info:
https://encrypted.pcode.nl/blog/2013/04/14/display-profiles-generated-from-edid/
https://encrypted.pcode.nl/blog/2013/11/24/display-color-profiling-on-linux/

#10 Updated by Michel Corps almost 2 years ago

Pascal de Bruijn wrote:

Also keep in mind that Darktable applies any active display profile by default (unlike many other programs, who incorrectly ignore it by default), thus if you have a broken display profile loaded it might affect color rendering negatively as well.

In my case I have a calibrated screen (using a Color Munski device).

This "problem" occurs with different makers (Olympus, Canon, others ?) and only with a particular blue color.
Someone could try using RawTheerapee with such images ?

#11 Updated by Ilya Kruchinin almost 2 years ago

Hi Pascal,
In regards to:
"keep in mind that many modern desktops enable a socalled EDID profile by default these days, which is sourced from the display hardware"
I tried pretty much everything. * Disabled colord * Loaded standard color profiles * Loaded color profiles from other monitors * Tried editing the color profile manually
I just don't think it is related (well, with both "color profile" disabled in GNome and a completely disabled "colord" daemon it should have no affect, as EDID is not used for color profiles)
I did erase darktable metadata/cache as well between each trial - no luck.

I do understand that the colors get messed up once I click on "darkroom" tab - a new "preview" gets generated with broken colors.
I do, as you mentioned, suspect that some "magick trickery" with display color profiles might to be blame.
I could NOT find a way to disable color profile processing in darktable (there is no tickbox for that) to prove it, however.

An another note, however. When I first import pictures to darktable - the previews look fine in "lighttable" mode. It's only when I FIRST click on "darkroom" when colors become corrupted and are then written in corrupted form to a cached preview (e.g. will, from then on, be displayed incorrectly in the "lighttable" mode).

Can I assist you in providing any additional information? What could shed more light on the issue?

#12 Updated by Michel Corps over 1 year ago

Ilya Kruchinin wrote:

An another note, however. When I first import pictures to darktable - the previews look fine in "lighttable" mode. It's only when I FIRST click on "darkroom" when colors become corrupted and are then written in corrupted form to a cached preview (e.g. will, from then on, be displayed incorrectly in the "lighttable" mode).

That seems to me a normal behaviour ; DT uses first the RAW embedded preview which is built by the camera with a "right" blue.

#13 Updated by Michel Corps over 1 year ago

To recapitulate :

The problem oocurs only with a particular blue.
it occurs whith different makers (Olympus and Canon and my case).
It occurs with base curve disabled and/or gamut clipping enabled.
It occurs with a calibrated screen.

I just tried with RawTherapee : no problem. I don't know from where RT get its color profiles but we can assume there is a similar process to get them from Adobe. So is it really impossible to imagine that there is a bug in the DT process to elaborate this color profiles from Adobe ?

#14 Updated by Roman Lebedev over 1 year ago

Michel Corps wrote:

To recapitulate :

The problem oocurs only with a particular blue.
it occurs whith different makers (Olympus and Canon and my case).
It occurs with base curve disabled and/or gamut clipping enabled.
It occurs with a calibrated screen.

I just tried with RawTherapee : no problem. I don't know from where RT get its color profiles but we can assume there is a similar process to get them from Adobe. So is it really impossible to imagine that there is a bug in the DT process to elaborate this color profiles from Adobe ?

dt indeed by default uses camera color matrix from adobe. and that is the only thing that dt uses from adobe.
i think, rt uses DCP, which have lots more info, but are also kinda non-free. dt does not use DCP and will not use them.

Expectation that one can get the same exact output of different software is false. every program has it's own way of doing things.
If you really care about color, i'd recommend getting it8 target, and following https://www.darktable.org/2016/05/colour-manipulation-with-the-colour-checker-lut-module/
(and maybe create custom color matrix, but that is harder)

#15 Updated by Michel Corps over 1 year ago

Expectation that one can get the same exact output of different software is false. Every program has it's own way of doing things.

I am not talking about to get exactly the same color as the camera jpeg or another software. I tried RT just because I supposed it get color profile from Adobe as DT, not to check it produces exactly the same colors. The "problem" is a particular color which is wrong (oversatured) when the others colors are only slightly different from the camera jpeg (or others raw processors).

I don't really care about color (it is a bit to complicated to me) but when this wrong blue occurs it is impossible to export the jpeg.

I get a dress producing the problem so I will do some others tries with my differents cameras.

#16 Updated by Roman Lebedev over 1 year ago

Roman Lebedev wrote:

dt indeed by default uses camera color matrix from adobe.

Michel Corps wrote:

Expectation that one can get the same exact output of different software is false. Every program has it's own way of doing things.

I tried RT just because I supposed it get color profile from Adobe as DT, not to check it produces exactly the same colors.

color matrix != color profile
color profile == color matrix + curves + ??? + some more half-proprietary stuff

#17 Updated by Michel Corps over 1 year ago

Roman Lebedev wrote:

color matrix != color profile
color profile == color matrix + curves + ??? + some more half-proprietary stuff

Oh... right. As as said I am not very expert in this domain. Thanks for the correction.

#18 Updated by Ilya Kruchinin about 1 year ago

Just an update.
I found the root cause that's messing up blue color.
It's the CAMERA color profile (NOT the monitor color profile and NOT the base curve).
For some reason, starting with a certain version of darktable the blue colors for 600D became all wrong.

I was able to partially resolve it by installing a camera profile provided by other people (I don't have a color target to create a profile myself).
I got the camera profiles from here: http://darktable-users.narkive.com/veqty7ir/wrong-colours-with-canon-60d-too-much-blue and tried a couple.
None of them are perfect, but at least by using them I can get a decent image to export.

I'd really much prefer if this (camera profile) could be fixed in darktable without using any third-party files.
If you need any additional info - please let me know.

Also available in: Atom PDF