PanoTools 2.2: Pincushion and CA correction

Al Evans · Aug 4, 2003

I spent a good deal of yesterday afternoon fiddling with a set of highly rectilinear test images I shot, looking for the right settings for the PanoTools-> Correct-> Radial Shift filter to get rid of pincushion distortion and color fringing.

Both of these kinds of distortion seem to be pretty bad at the widest-angle setting, and minimal at full telephoto (but the pincushion distortion never seems to change to barrel distortion).

And let me assure you, that was as DULL a piece of work as I've ever done! Fortunately, you shouldn't need to repeat it. These problems are a function of the lens design, so my results should work for you, too.

Theoretically

I believe that the lens is the same on E10s and E20s, so these parameters should work for either camera.

I shot the test images with the lens of my E20 wide open, at each marked zoom setting. I made them REAL BIG in Photoshop, and tweaked the Radial Shift filter settings until I was satisfied that straight lines were straight and there was no color fringing out at the edges.

Disclaimer: I was unable to find a solid statement that these kinds of distortion are independent of lens opening, and I don't remember for sure. But I THINK they are.

Here are the settings I ended up with, for each marked focal length:. I apologize for all the hyphens -- not only do tabs not work right here, even spaces seem to disappear:

9mm
------------d-----------c-----------b-----------a
R-------1.01851-----0----------0.018------0
G-------1.018--------0----------0.018------0
B-------1.0185-------0----------0.018------0

13mm
------------d-----------c-----------b-----------a
R-------1.00835-----0-----------0.008-----0
G-------1.008--------0-----------0.008-----0
B-------1.00835-----0-----------0.008-----0

18mm
------------d-----------c-----------b-----------a
R-------1.0049------0----------0.0045-----0
G-------1.0045------0----------0.0045-----0
B-------1.0049------0----------0.0045-----0

25mm
--------------d-----------c-----------b-----------a
R-------1.00415------0-----------0.004------0
G-------1.004---------0-----------0.004------0
B-------1.00415------0-----------0.004------0

(I did this last one just for completeness. There isn't really any noticeable distortion of either kind. I would probably never actually use it.)

36mm
------------d-----------c-----------b-----------a
R-------1.0015------0-----------0.001------0
G-------1.001-------0------------0.001------0
B-------1.001-------0------------0.001------0

--Al Evans

Al Evans · Aug 4, 2003

Al Evans said:
...pincushion distortion and color fringing.

Well, dagnabbit, I meant barrel distortion.

I never CAN remember which of those is which, even right after I look it up.

--Al Evans

Doehns · Aug 4, 2003

Did you try ptlense from Thomas Niemann yet? Does a good job on distortion.
http://epaperpress.com/ptlens/index.html

And you could compare your settings to his profiles for the E-10/ E-20

oh, and it's Freeware.

PDM · Aug 4, 2003

Thanks Al for going through the details. I'm sure this will be much appreciated next time I stich a pano. Any chance your going to do the same for the WCON at 9mm?

Paul

Al Evans said:
I spent a good deal of yesterday afternoon fiddling with a set of
highly rectilinear test images I shot, looking for the right
settings for the PanoTools-> Correct-> Radial Shift filter to get rid
of pincushion distortion and color fringing.

Both of these kinds of distortion seem to be pretty bad at the
widest-angle setting, and minimal at full telephoto (but the
pincushion distortion never seems to change to barrel distortion).

And let me assure you, that was as DULL a piece of work as I've
ever done! Fortunately, you shouldn't need to repeat it. These
problems are a function of the lens design, so my results should
work for you, too.

Theoretically

I believe that the lens is the same on E10s and E20s, so these
parameters should work for either camera.

I shot the test images with the lens of my E20 wide open, at each
marked zoom setting. I made them REAL BIG in Photoshop, and tweaked
the Radial Shift filter settings until I was satisfied that
straight lines were straight and there was no color fringing out at
the edges.

Disclaimer: I was unable to find a solid statement that these kinds
of distortion are independent of lens opening, and I don't remember
for sure. But I THINK they are.

Here are the settings I ended up with, for each marked focal
length:. I apologize for all the hyphens -- not only do tabs not
work right here, even spaces seem to disappear:

9mm
------------d-----------c-----------b-----------a
R-------1.01851-----0----------0.018------0
G-------1.018--------0----------0.018------0
B-------1.0185-------0----------0.018------0

13mm
------------d-----------c-----------b-----------a
R-------1.00835-----0-----------0.008-----0
G-------1.008--------0-----------0.008-----0
B-------1.00835-----0-----------0.008-----0

18mm
------------d-----------c-----------b-----------a
R-------1.0049------0----------0.0045-----0
G-------1.0045------0----------0.0045-----0
B-------1.0049------0----------0.0045-----0

25mm
--------------d-----------c-----------b-----------a
R-------1.00415------0-----------0.004------0
G-------1.004---------0-----------0.004------0
B-------1.00415------0-----------0.004------0

(I did this last one just for completeness. There isn't really any
noticeable distortion of either kind. I would probably never
actually use it.)

36mm
------------d-----------c-----------b-----------a
R-------1.0015------0-----------0.001------0
G-------1.001-------0------------0.001------0
B-------1.001-------0------------0.001------0

--Al Evans

Al Evans · Aug 5, 2003

PDM said:
Thanks Al for going through the details. I'm sure this will be much
appreciated next time I stich a pano. Any chance your going to do
the same for the WCON at 9mm?

I'll be happy to! Just send me the WCON!

Seriously, though, I don't even do panoramas. It just drives me nuts for straight lines to be crooked.

--Al Evans

Manfred Helsper · Aug 5, 2003

Al Evans said:
PDM said:

Thanks Al for going through the details. I'm sure this will be much
appreciated next time I stich a pano. Any chance your going to do
the same for the WCON at 9mm?

Click to expand...

I'll be happy to! Just send me the WCON!

Seriously, though, I don't even do panoramas. It just drives me
nuts for straight lines to be crooked.

--Al Evans

--Al,

Excuse me for not knowing what these letters over your data mean. Can you explain. Also, can one draw any conclusions from your data? Thanks,
Manfred

PDM · Aug 5, 2003

I'd have to agree with you there. Lens distortions and misregistrations really bother me too. I need to remember to run all my pics through this for correction. Now about the settings for the WCON...

Paul

Al Evans said:
PDM said:

Thanks Al for going through the details. I'm sure this will be much
appreciated next time I stich a pano. Any chance your going to do
the same for the WCON at 9mm?

Click to expand...

I'll be happy to! Just send me the WCON!

Seriously, though, I don't even do panoramas. It just drives me
nuts for straight lines to be crooked.

--Al Evans

Al Evans · Aug 5, 2003

Manfred Helsper said:
--Al,
Excuse me for not knowing what these letters over your data mean.
Can you explain. Also, can one draw any conclusions from your data?

Certainly.

Briefly, the "Correct" filter in Panotools uses an equation with four coeffficients to remap the pixels (thus correcting radial distortion. The letters (a, b, c, d) are the names of the coefficients. Here's a pretty good explanation:

http://www.caldwellphotographic.com/TutorialsDistortionAndColorFringing.html

The only real "conclusion" that could be drawn from the data is that both sorts of distortion are less at longer focal lengths -- but this is what we would expect

The usefulness of the data lies in the fact that it should work for all E20 (and, presumably, E10) lenses -- so you can use my numbers and avoid a couple hours' work.

--Al Evans

Manfred Helsper · Aug 7, 2003

Al Evans said:
Al Evans said:

--Al,
Excuse me for not knowing what these letters over your data mean.
Can you explain. Also, can one draw any conclusions from your data?

Click to expand...

Certainly.

Briefly, the "Correct" filter in Panotools uses an equation with
four coeffficients to remap the pixels (thus correcting radial
distortion. The letters (a, b, c, d) are the names of the
coefficients. Here's a pretty good explanation:

http://www.caldwellphotographic.com/TutorialsDistortionAndColorFringing.html

The only real "conclusion" that could be drawn from the data is
that both sorts of distortion are less at longer focal lengths --
but this is what we would expect

The usefulness of the data lies in the fact that it should work for
all E20 (and, presumably, E10) lenses -- so you can use my numbers
and avoid a couple hours' work.

--Al Evans

Thanks Al for your explanation.
Manfred

Search

PanoTools 2.2: Pincushion and CA correction

Al Evans

Forum Enthusiast

Al Evans

Forum Enthusiast

Doehns

Senior Member

PDM

Senior Member

Al Evans

Forum Enthusiast

Manfred Helsper

Active member

PDM

Senior Member

Al Evans

Forum Enthusiast

Manfred Helsper

Active member

Keyboard shortcuts