Auto matched curve misbehaves on this shot

[Former user]

Left is “Standard Film Curve”, right is “Auto matched curve”.
Tested with ART recent git master, RT-5.6 and RT-git [current newlocallab].

Raw is here: https://discuss.pixls.us/uploads/short-url/zRtrBWtI83zkfygDZ3H44qredhn.RW2

Originally reported for RawTherapee: https://discuss.pixls.us/t/rt-issue-with-auto-matched-curve-with-this-shot/13629

[agriggio]

i had a quick look. it could be possible to get a better match for this shot, but at the cost of significantly complicating the logic, with the risk of introducing regressions for other pictures. not worth it IMHO. the auto matching is necessarily heuristic and it will never be perfect. so I'm not going to do anything about this, but I'll keep the issue open in case some other ideas pop up

[agriggio]

I have pushed a couple of changes that should improve the robustness of histogram matching. Let me know if they help…

[Former user]

Thank you for the improvement.

I tried many old photos developed with auto matched curve and didn’t find a breakage. However I found some more photos that don’t look like the OOC jpeg. Some of them are ISO16000 shots with B+W setting, the others include my Labrador. I think I will stop here. In any of the photos (including the one in this issue) standard matched curve looks close enough. If you want I still can send you some of those files. Otherwise just close the issue.

Thanks!

[agriggio]

A perfect match will in general be impossible, especially if you use custom “looks” in the camera, which might be based on 3d LUTs, not a simple tone curve. However, if you have some pictures that are way off (especially in an unpleasant way), feel free to send them to me and I’ll take a look.

In the meantime, you can close this if you are happy with the results.

[Former user]

These photos I found only showed low contrast which could easily be fixed. If you want to have a look here is just one sample:

https://www.dropbox.com/s/njtmdnnzeu9esk7/P1080442.RW2?dl=0

[agriggio]

Thanks for the sample. I had a quick look but I see nothing simple that I can do to improve the matching in this case. I will therefore close the issue for now