Alberto, thanks a lot for your good will :), but unfortunately it is still far from perfect. It looks like I need to hurry up with leaving Windows XP ;), but probably you should also take a closer look at this issue.
I think there was a misunderstanding. the “fixed” LCP file I provided was meant to be good only for your first pic. that is because I left only one entry in it, the one that should be good for your previous pic. it is not surprising that it doesn’t work for other pictures. it was not meant to be a solution, the purpose was to demonstrate what goes wrong in the current version of RT.
btw, I did take a closer look, in fact I provided a (experimental) patch that tries to fix the issue. but since I am not seeing those problems on the lenses I use, the only hope to see the patch getting integrated into a future RT release is for users to help and test it (hint…) that means that you will have to compile RT yourself (probably in a virtual machine if you are stuck on XP). I know this can be annoying, but I have no Windows machine so I’m afraid I can’t help here…
@gaaned92 thanks! I suppose it can be helpful, but you need to ask the interested users. as I wrote, the current RT works fine for my lenses. I will do more testing to see that the patch doesn’t break the LCP support for the lenses I have, but I don’t have many this is really a case where we need help from the community!
I can ask for help on Google+ once we have a build (not gtk2). @agriggio can you define how best to test? What sort of info should the raw file have or not have?
I have changed the logic for finding the right distortion coefficients in the LCP files. basically we should look for two things:
no regressions. if you are using LCP for your lenses and you are happy with the results, please check that it is still the case with this patch
better results for problematic lenses. if you have a lens with an LCP that is supposed to work (e.g. coming from Adobe) but RT does a poor job with it, please check if the situation improves with this patch.
in both cases, if possible test with as many apertures, focal lengths and focus distances as possible.
I agree that lensfun solution would be better, still about lcp handling in this github issue there is a link to a previously “bad” lcp and an image that showed problems, together with the exported jpeg with a previous rt version: Lens distorsion correction · Issue #2944 · Beep6581/RawTherapee · GitHub
Unfortunatly i’m still using ubuntu 14.04 so i can’t compile from latest, but maybe it could be interesting to try.
A compiled test version for Windows (x64. Gtk3.22-15) can be downloaded from here:
No installer included. Extract the folder “RawTherapee_lcp” to e.g. your Desktop and run “rawtherapee.exe” inside this folder.
Cache and settings are saved into “localappdata\RawTherapee-lcp”. It leaves your existing installation untouched.
update: I’ve tweaked the matching logic a bit more, essentially copying what lensfun is doing when converting LCP profiles. although unfortunately the pic of @dvan is still not properly corrected (now it undercorrects inserted of overcorrecting), the new logic makes more sense than my previous attempt and it seems to be equally good for the other problematic pictures that have been posted here.
so, I’d like to ask all interested people to test this new version. if there are no regressions, I’d like to merge this asap so that it is included in 5.2 and I can move on to integrating lensfun
@TooWaBoo thanks for the quick turnaround. I will revert to the previous method then! if no more issues arise, I’ll try to merge in a couple of weeks (just to see whether something else pops up in the meantime)