I have the Samyang 12mm wide-angle lense, which is listed as supported in lensfun for APS-C camera, but not my Olympus OM-D E-M10 Mark II.

According to this article, I need some kind of diffuse material, but I don't have white plastic on hand. I do have sheets of paper though.

Does anyone have experience with performing this kind of calibration? Are there other household materials that you would recommend instead of sheets of paper?

  • 1
    Regardless of what material is used, the critical factor for getting accurate results is to insure that the material is evenly illuminated by light sources that illuminate the entire surface of the material at the same intensity.
    – Michael C
    Oct 31 '17 at 21:36
  • Do you have access to some lightbox used to view slides or negatives? They should bei illuminated rather evenly.
    – Gerhardh
    Oct 31 '17 at 21:53
  • 1
    BTW, I added a few mounts, including Micro 4/3, for this lens in lensfun, so the existing profile should work going forward. Nov 15 '17 at 20:39

I do have some experience with the particular calibration tool in question, and while it's hard to give a general answer to questions like "Is paper an ok diffuser?" (there are many kinds of paper), I'll share my experience.

Firstly, I feel a little differently about the importance of calibrating vignetting at different focus distances than what Torsten expresses in his tutorial. My experience with the lenses that I own is that they have stronger vignetting when focus is at infinity than at closer distances. The implication of this is that if you only calibrate at infinity, close shots will be over-corrected.

In my opinion, over-correction of vignetting looks terrible, both because it looks very unnatural compared to the "problem" that it's supposed to be correcting, and because it tends to pull up noise from shadows in the corners (where the correction can typically be as much as 2 stops for wider lenses). I gathered from personal correspondence that the reason for downplaying the importance of multiple distance calibration in the tutorial is in hopes of making the task less overwhelming to anyone considering it, but I would say that it's well worth doing at least MFD and something around 2-6 times that, unless the vignetting is clearly very similar at both ends of the focus range. If you really want to only take images at one focus distances (perhaps for a manual-focus lens that doesn't produce EXIF information about distance, or a camera system where the distance reporting is unreliable) it's probably best to calibrate for the least case, wherever in the range it occurs.

Now as to the actual question about methods of getting an evenly lit calibration shot... it's complicated. I've gotten good results using the diffuser-in-front-of-the-lens method for some lenses, and not others. In general, it seems like the wider-than-normal lenses can produce calibrations that, when applied to real-world images, result in the aforementioned horrible icky over-correction. I don't know exactly why this is the case, but for those lenses it was true for all the different diffusion methods I tried. The main point being: check your results on some real-world images that will easily show this, at different focus distances.

In the cases where this method did give good results, these were obtained by making sure that the diffusion material was held tightly against the face of the lens by something flat (such as glass), so that no light can "leak" at the edges. Eventually when I obtained a device with an OLED display I found that setting the lens on a sheet of diffuser on top of a full white screen worked as well as anything I had tried previously, but DO NOT do this with backlit screens - they are not reliably evenly illuminated. Otherwise, the diffuser can be lit as described in the tutorial, by pointing at an evenly lit ceiling. Extremely ambient lighting is best, it doesn't need to be bright, long exposures are fine...

...which brings me to the part about those lenses that didn't seem to give good results with this method. For those, I ended up pointing the lens, without a diffuser, at a very dimly, evenly illuminated ceiling, such that the exposures lasted a few seconds, and rotated the lens around it's axis during the exposure to further average out any uneven lighting, trying for roughly one full constant rotation during the exposure. This turned out to work quite well, and the dizziness wore off eventually. :D

As for the diffusing material, I tried various things I had around, which included some Roscolux diffusion gels. That stuff works great if it's in pristine condition, but it can easily get scuffed in ways that make it unsuitable. I've also found that multiple layers of LDPE sheet (available in many craft stores) can also give very good diffusion. Probably some types of paper can, as well, but some are surprisingly "blotchy" when held up in front of a light. The main thing is to check whatever you're considering using this way, to make sure it isn't contributing any uneven distribution of the light. I would probably also reject light tables as a direct light source, since they probably don't have the critically even illumination needed (taking a photo of it might reveal this).

So, that's my brain dump. One other thing worth mentioning is that the script depends not only on python, but also some hairy math libraries that you'll probably need to install somehow, so be prepared for that.

ADDENDUM: I had a reminder today about how image stabilization mechanisms can produce variations in vignetting. This needs to be disabled for the shots.


Haven't tried this type of calibration - but it seems that a sheet of wax paper would work, single sheet of printer paper, or some thin white cotton as well.

That being said - the lens can cover up to APS-C, which is a bit larger than the micro four thirds sensor in your Olympus. Any apparent vignetting at APS-C will be minimized in your shots due to simply being cropped out of the captured image circle.

Given that, I have to ask, why bother going through the trouble? Vignettes are sometimes nice to have and they're easy enough to fix in post where you don't want them.

Edit to add: As has been pointed out - the object of use isn't terribly important - even illumination is. Even illumination can be provided easily by using an overcast sky. I imagine a blue sky with the sun behind you would probably work as well.

There's a discussion about it here: DIY Vignette Measuring

  • I figure if I'm going to bother to get the calibration for distortion and chromatic aberration, I might as well try and get the vignetting, even though, like you said, the lens is built for APS-C, so vignetting is probably minimal. I can't remember if darktable had the option of choosing what kind of correction I would like to apply...
    – Calyth
    Oct 31 '17 at 15:17
  • "Vignettes are something nice to have" a rather questionable or at least highly opinion based statement.
    – Gerhardh
    Oct 31 '17 at 21:48
  • 1
    @Gerhardh - "sometimes" is the chosen word. As in, "sometimes" they contribute to the photo, other times they detract. Since the statement is so ambiguous as to include all instances of vignette from none to full and is also ambiguous in judgement of those vignettes, it can be neither questionable nor opinion based. It is, in fact, a fact. Unless of course, you want to argue that vignettes are never nice to have...which, in fact, would be an opinion.
    – OnBreak.
    Oct 31 '17 at 22:36
  • @Calyth Darktable allows you to apply distortion, TCA, and vignetting correction in any combination. Nov 1 '17 at 1:08
  • Awesome @junkyardsparkle. I do find vignetting useful to highlight the subject, so sometimes I even apply the effect in dark table.
    – Calyth
    Nov 1 '17 at 1:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.