PTGui to stich scanning electron microscope images?

[Bjoern Ludwar]

I am trying to find a way to stich images taken with a scanning electron microscope. Commercial solutions cost $$$$ - more than our school can afford und I *think* PTGui might be able to do it.

Here is the issue: Unlike with a panorama where the camera rotates around a fixed point, in the SEM the objects is shifted under a fixed SEM objective lens (akin to aerial photography) - so the camera position isn't identical for each image. However, unlike the flat images obtained with a light microscope, the SEM's images show perspective distortion - which is why widely available (light) microscopy stitching programs won't work. 

There is no sensor size, or known focal length, so I have been playing around with every combination of parameters I can think of - PTGui gets a subset of images assembled correctly, but then fails to do the rest. Here is a sample set (12 partial images with 50% overlap - 3 columns and 4 rows, naming is sequential starting in the upper left corner). Any ideas/thoughts/suggestions as to how to assemble those into a panorama (if that is possible with PTGui)?

Thanks and best, Bjoern

[Erik Krause]

Am 22.11.2022 um 22:31 schrieb Bjoern Ludwar:
> Here is the issue: Unlike with a panorama where the camera rotates around a
> fixed point, in the SEM the objects is shifted under a fixed SEM objective
> lens (akin to aerial photography) - so the camera position isn't identical
> for each image. However, unlike the flat images obtained with a light
> microscope, the SEM's images show perspective distortion - which is why
> widely available (light) microscopy stitching programs won't work.

They don't only show perspective distortion, they also show heavy
parallax. For a flat approach I would have recommended a workflow like
described in https://ptgui.com/support.html#6_5
However, parallax makes things far more complicated if not impossible.
In fact parallax is so huge, I'd rather suggest some 3D-reconstruction
software. You should find quite a few - even open source - programs if
you search for "structure from motion".

The images are awesome, by the way...

--
Erik Krause
http://www.erik-krause.de

[Misho Ristov]

I’d suggest to use a photogrammetry software like

https://www.capturingreality.com/

and export an ortho photo from the scan. or use the 3d model.

You’d need to shoot with more overlap though.

*not affiliated with the soft, just a long time user.

cheers,

Misho

--
You received this message because you are subscribed to the Google Groups "PTGui Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ptgui+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ptgui/cf6c67dc-6738-40e7-ae1d-d19399bdcf73n%40googlegroups.com.

[John Houghton]

On Tuesday, November 22, 2022 at 11:06:12 PM UTC Erik Krause wrote:

They don't only show perspective distortion, they also show heavy
parallax.

Always up for a challenge  -  I tried using viewpoint correction and got a result of sorts, but heavily distorted!

John

[Bjoern Ludwar]

Thank you for all the answers and for pointing me in the right direction - I'll pursue "structure from motion" / 3D reconstruction. 

John: thank you especially for putting in the time and trying the viewpoint correction. I had hoped that PTGuiPro might be my ticket, but it looks like it is not quite the right tool for the job (though I like the panorama for non-science purposes!).

I truly appreciate all of your help!

Bjoern