Thanks for the reply junkyardsparkle. I thought my previous message
disappeared b/c it didn't show up in the archive.
For the vignetting calibration I tried stretching a thin piece of a white
plastic bag over a glass plate. The images look like plain blank gray
rectangles with vignetting, which seems right.
I'm a bit suspicious of the calibration values I get. When I apply the
correction to the original test image, the corrected version has a "halo"
in it, a brighter circle a little ways in from the corners. I only see it
in a thumbnail-sized version of the image. Viewed at full screen (which is
still a small fraction of the image's full size) it looks ok. Maybe this is
normal? For my purposes it is good enough, and far better than the
uncorrected version. I can't see any problems with photos of normal scenes.
But if I'm going to submit the calibration data for inclusion in the
LensFun database I feel I should make sure it is as good as possible.
Am I correct to think that a good calibration image should have brightness
values that are symmetric? In other words, if I compare my image to a copy
that has been flipped over the horizontal axis, they should look the same?
I checked this using matlab, and the largest difference between the
original and flipped image (R,G,B) is (18,12,11) on a 0-255 scale. The sky
was completely clear and the sun was going down. I aimed at a "plain" blue
section, but it's possible there was a gradient.
You make a good point about focus breathing. I think I'll wait for an
overcast day (to get more even light) and then re-shoot with multiple focus
distances.
Thanks again,
Owen
On Sat, Apr 16, 2016 at 8:10 PM, <[email protected]> wrote:
>
> On Sat, Apr 16, 2016, at 15:16, Owen Mays wrote:
>
> > It seems my previous attempt to email this list disappeared. I apologize
> > this shows up twice!
>
> It showed up for me, but I assumed somebody more qualified would respond.
> Since that hasn't happened yet, I'll comment based on my own experiences,
> limited to a couple of lenses.
>
> > For Vignetting correction, I read Torsten's instructions here:
> > http://wilson.bronger.org/lens_calibration_tutorial/
> >
> > How even does the even illumination need to be? If I have a suitable
> > diffuser, could I aim at the sky for the vignetting images? I am
> concerned
> > that if I point a standard floor lamp at the ceiling it will have a
> > distinct circular illumination pattern.
>
> As even as possible, but you can also mitigate variation somewhat by
> taking multiple exposures while rotating your camera/lens axially... I do
> three at ~120-degree intervals. Also, although the advice on that page says
> that exposures at multiple focus distances aren't terribly important for
> most lenses, I've found that it matters very much for one of mine (I'm
> about to do a third, more complete series of calibration shots right now,
> in fact). I suspect this is at least somewhat dependent on how much the
> (actual vs. nominal) focal length of the lens "breathes" with focus
> changes, in general, but that's just a guess. It's easy enough to take a
> series of shots to test for a given lens.
>
> Probably the most important thing to watch for is that the diffusing
> material fits perfectly against the front of the lens tube with no gaps,
> since light leaking in could potentially skew the results as much as uneven
> lighting, maybe more so. I like to visually sanity check my calibration
> images before running the script on them; rotating them by 90-degree jumps
> in your viewer makes it easier to detect any odd variations.
>
> > For TCA correction: I haven't found a good target for TCA. Does this
> > calibration need a regular geometry (like the buildings used for
> > distortion?) Or are trees against an overcast sky ok?
>
> Regularity isn't important, you just need sharp contrasty (but not
> overexposed) non-radial lines that are fairly near the outermost parts of
> the image. Diagonal lines in corners will probably show the most detectable
> TCA. Your mileage may vary, but I've found that for my purposes, TCA also
> changes with focus distance, and since lensfun doesn't handle this, I don't
> enable it by default, since it's wrong as often as it's right. If you
> generally focus at similar distances with a given lens (portrait,
> landscapes, etc.) this might be less of an issue, and some lenses may not
> exhibit this behavior for all I know.
>
> For whatever that's worth. :)
>
> --
> junkyardsparkle
>
>
>
> ------------------------------------------------------------------------------
> Find and fix application performance issues faster with Applications
> Manager
> Applications Manager provides deep performance insights into multiple
> tiers of
> your business applications. It resolves application problems quickly and
> reduces your MTTR. Get your free trial!
> https://ad.doubleclick.net/ddm/clk/302982198;130105516;z
> _______________________________________________
> Lensfun-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/lensfun-users
>
------------------------------------------------------------------------------
Find and fix application performance issues faster with Applications Manager
Applications Manager provides deep performance insights into multiple tiers of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial!
https://ad.doubleclick.net/ddm/clk/302982198;130105516;z
_______________________________________________
Lensfun-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/lensfun-users
