Menu
▾
▴
#113 Sony E PZ 16-50mm F3.5-5.6 OSS vignetting
The Sony E PZ 16-50mm F3.5-5.6 OSS vignetting data is off. There is a an example of this issue here:
https://discuss.pixls.us/t/lensfun-vignetting-correction-for-sony-kit-lenses/9537
I think the issue is more pronounced with real world images. If I find time I'll try to create a corrected profile.
Related
Discussion
I suspect some slight over-correction might be a common issue with some of the profiles for wider lenses in the database. Can you confirm that the results are the same using the lens correction in darktable? If so, I'll add a FIXME comment to the profile in the db.
So far I only used the lens correction within darktable. I was under the
impression that darktable just applies the lensfun calibration profiles. Do
you want me to test the correction profile without darktable? Or what kind
of confirmation do you want?
I could easily provide a real life example, that I corrected with the
supplied lens profile of the darktable windows installer. The
over-correction of the vignetting is very pronounced. Would this help you?
Regarding the recalibration: I already bought a sheet of acrylic, but
haven't found the time yet to take some pictures and send them towards the
calibration script.
Am Di., 29. Jan. 2019 um 03:07 Uhr schrieb junkyardsparkle junkyardsparkle@users.sourceforge.net:
Related
Bugs: #113
If you say you notice it in darktable, that's confirmation enough... I only ask because it wasn't clear in the linked discussion what software was being used, and darktable's implementation is the one I know. :)
If you think you might get around to creating new data fairly soon, I'll just wait for that rather than adding a comment to the profile.
Today I took a closer look at the vingetting problem. I took a picture at 16mm f/3.5 and infinity focus with the acrylic sheet in front of the lens. To my eyes the corrected image looks perfect and even better than the camera corrected one.
Then I compared this result with two real world examples (also 16mm, f/3.5). In these pictures there appears to be some overcompensating (at least to my eyes).
Any opinions on this? I attached the images to this post and I'm happy to share the RAW files.
Update: I uploaded a wrong file. Now the effect is even more pronounced.
Last edit: Felix Vollmer 2019-02-10
This doesn't surprise me. The images used for calibration should always look well-corrected - the correction model itself does a really good job with normal, radial light fall-off (it doesn't correct color vignetting, though, so that may be apparent even in the calibration images). The problem is that it's hard to produce really good, evenly illuminated calibration images. The diffuser-directly-on-lens method is easy, but my experience is that, for wider lenses, it actually fails for some reason (by increasing apparent vignetting and producing over-correction on real-world images). There's some discussion about that here.
Maybe it's the focus distance, that changes the vingetting? According to the exif data ("Focus Distance 2") the distances are:
DSC05492: 37,93 m
DSC05500: 26.82 m
DSC06489: inf (acrylic sheet)
I could try to take a picture with acrylic sheet at a closer focus distance
Last edit: Felix Vollmer 2019-02-10
Focus distance is very much a factor for the modern internal-focus lenses I've used (all Micro 4/3, so compactness of design could play a part, I suppose). I wouldn't expect it to be a major factor until the distances were smaller than the ones in those pictures, but I think it probably depends on the specific lens in question, and might be hard to make general predictions about (strangest case I've seen so far is this one).
For this reason, I feel somewhat differently than Torsten Bronger about the wisdom of only profiling lenses at infinity. For unit-focusing lenses this is probably good enough, but for lenses with moving elements I think it's a bad idea, because over-correction at closer distances seems likely, and it looks much worse than under-correction. The concern is that making the profiling too complicated will prevent people from doing it, but I think people willing to do it at all are probably looking for good results, so... I don't know what the best compromise is.
I would say that if you only want to profile at one focus distance, then it should be the distance that produces the least vignetting, so that it errs on the side of under- rather than over-correction for the rest of the range. Indeed, much in-camera correction seems like it may use this approach, as most images only have partial correction... this may also be an esthetic choice; absolute correction isn't always what people want to see, unless they're stitching panoramas. Of course, determining this distance might end up not being less work that just taking calibration shots at multiple distances... but given that there are still many cameras around that don't produce reliable distance info in EXIF, this may still be the best general, all-around approach.
Another intresting fact is that the Vignetting Corr Params, which are embedded inside of the files, are different:
DSC05492: 0 64 256 640 1280 2112 3392 5440 8064 11264 14912 0 0 0 0 0
DSC05500: 0 64 256 640 1280 2112 3392 5440 8064 11264 14912 0 0 0 0 0
DSC06489: 0 64 256 640 1280 2112 3456 5504 8192 11392 15168 0 0 0 0 0 (the acrylic sheet in front of the lens with focus at infinity)
That might be particularly interesting to the people trying to reverse-engineer those parameters. :D
I did those calibrations myself back in 2012 because I own those lenses. I had a look at my old images – the correction in Darktable is perfect. Thus, there are three possible explanations:
The second point can be excluded easily: Just assure that in-camera vignetting correction is off. At least in many Sony models, it is also applied to RAWs!