Strange effects with saturation in colour balance RGB and filmic

Processing
1

Using development build 3.5.0+2159~g22aba1f90 for windows 10, I have noticed strange artifacts when colours go out of gamut.
If I increase the saturation in colour balance RGB, the out of gamut colours turn black. This only happens when colour preservation in filmic is set to no, it doesn’t happen when it is on the other settings such as power norm. I saw a similar issue on github
https://github.com/darktable-org/darktable/issues/9095, it looked like it could have been fixed, but I wasn’t sure. I know that it is better not to push the saturation so things go out of gamut, but behaviour looks like a calculation tripping up. I noticed with the colour balance module in 3.4.1.1 too.

image
image1149×716 224 KB

image
image1135×710 248 KB

Borth 58.RW2 (14.7 MB)
Borth 58.RW2.xmp (11.2 KB)

This file is licensed CC BY-NC 4.0.

2

3.5.0+2516~g859a6adf8

I think so…mostly…maybe my test image…all settings are good up to 100% except chroma At 39% it goes black and then bleeds some red…

image
image1383×803 151 KB

image
image1399×814 160 KB

image
image1402×799 146 KB

image
image1405×818 123 KB

-100%

image
image1393×808 116 KB

image
image1402×820 112 KB

image
image1404×809 25.9 KB

3

OK thanks, no need for me to report it then.
It behaves OK with the other chroma preservation modes in filmic so I guess you can work around it if they haven’t managed to fix it completely.

4

I think that’s how it’s supposed to behave. Saturation in Color Balance RGB is not “real” saturation. That’s chroma.

5

I’ve noticed this too, but assumed it was something i was doing wrong. It seems to have appeared soon after 3.5.0~git2054.2013d1dc97-1 which doesn’t show this effect (on my laptop) so I am still using that older version.

Linux mint 20.1, kernel 5.8.0-55-generic, Intel iris plus graphics.

6

That’s interesting. It happens with saturation control in colour balance in 3.4.1.1 on Windows too, so perhaps introduced earlier than that?

7

I’ve just played with your image (not mine) and I get the same problem (non-problem) as above, even with 3.5.0~git2054.2013d1dc97-1 .

8

OK thanks, is it any help to upload the files onto the post on github mentioned above or is that just duplicating what has been said already?

9

Usually the posts here are not seen as a bug reports but may lead to bug reports if discussed enough. In this case it’s the extreme saturation issue that might be fixed by colorbalancergb gamut mapping : better fix for #9095 by aurelienpierre · Pull Request #9227 · darktable-org/darktable · GitHub but if you don’t put your sample in it it might be missed.

If you can - check the mentioned PR if it’s not yet merged. If it fixes your issue - great. If you can’t check or confirm that it does not fix your issue then mention your sample in the issue.

1 Like
10

Funny up above I thought the chroma slider at -100 would be grey??

11

It is worth mentioning that dt issues are set to close automatically. I have had a few that expired without attention.

12

Off topic to the topic at hand but on-topic for dt gh issues: the bot message should “motivate” people to update the ticket with new info or ping devs or whatever. I try to check “least recently updated” issues to see whether they are still replicable on most recent master and if so - make sure they’ll stay open by adding new details or confirming that the present details are enough to work on the issue.

1 Like
13

quote from the official documentation: “Setting the global chroma to -100% will not produce a real monochrome image”

1 Like
14

Looks like this issue has been fixed now. Nice work!

15

I think I blasted the 200 working hours cap on this module…

6 Likes
16

Just a note, since the module author won’t self-advertise: everybody’s welcome to show some support via aurelienpierre's profile - Liberapay or paypal :wink: 200+h on module that’s damn near magic…

7 Likes
17

What’s the hour to slider conversion rate? :stuck_out_tongue: