Problem. "Fake rotation" of the object, with my Mitutoyo10x

[Medwar]

Hi all!

I have encountered a very strange problem.
When shooting with Mitutoyo 10x M Plan APO 0.28 on Lumix G9 camera.
When I change the focus by moving the camera to make the stack, the objects seem to be rotating between the frames!
Please check the GIF here, can't upload it directly to the forum. The resolution is low, but the problem is seen. There is a phenomenon, that looks like the rotation of the specimen in the (almost) vertical plane parallel to the line of shooting. The specimen stands stable.
https://makeagif.com/gif/rotate-mitu10x-8XBK89?

It looks like the parts in the out of focus areas get shifted - in this case behind the focus they are shifted up, in front of the focus - shifted down.

I am 95% sure that I have seen a topic on this forum with the same issue, maybe with a different lens.
But unfortunately I can't find it, spent almost an hour searching for it.

I made some tests, shooting the same object with the same lighting.
1) Tube lens change - effect is the same. Tried: Raynox-250, S-Componon 100mm, Sigma 70-200 EX.
2) Other objectives: Mitutoyo 5x, Mitutoyo 20x, Nikon 10x 0,25 - show no rotation, the stack is smooth and even.
3) When I rotate the Mitutoyo 10x in its thread, the plane of rotation changes accordingly. If I rotate Mitu 90 degrees, the rotation becomes horizontal from vertical.

My first conclusion was - I definitely have a defect Mitutoyo 10x.

But!
I have another copy of Mitutoyo 10x. And of course I tested it.
Rotation absolutely is the same! And moreover, if the other Mitutoyo 10x sits up to the same point, tightly screwed into the thread, then the orientation of the "fake rotation" plane for both objectives is the same!
It's hard to believe that 2 sharp Mitutoyos have acquired one and the same defect, not degrading in sharpness (stacks from both a bit sharper than Nikon 0.25, even considering that Mitu stack is not so smooth due to rotation).

Please, help me to clarify, what happens here?

[Scarodactyl]

Is it causing issues with stacking? If not I would not consider it a problem or worry much about what the out of focus areas are doing.

[Bob-O-Rama]

Its hard to tell, but sorta looks like an inversion of perspective as would happen in a hypercentric lens. But otherwise, yeah, this is the sort of counter intuitive stuff you see with 3D subjects.

[rjlittlefield]

Most likely your objectives are fine.

This sort of behavior is very common with 3D subjects and microscope objectives.

To explain... What happens with any lens is that the rendering of out-of-focus areas is determined by which light rays get through the lens to form the image. In most photography this only affects the smoothness of bokeh, depending on how the shape of the aperture interacts with edges in the subject and aberrations in the lens. But at high magnification with large NA, it becomes very likely that light preferentially enters one side of the lens for some parts of the subject. When this happens, the system acts just like the lens is looking "sideways" at those parts of the subject, so as the focus distance changes the apparent viewpoint does also. Many times the light preferentially enters different sides of the lens for different parts of the subject, which produces a particularly confusing "squirming around" effect like illustrated at viewtopic.php?p=149187#p149187 .

You may be able to make the effect less obvious by using illumination that strikes the subject more uniformly from all angles. But even that will not make it go away entirely whenever part of the subject is shaded or partly hidden by other parts.

--Rik

[CrispyBee]

I've had something like that happen when the camera wasn't properly aligned horizontally, meaning it was "dropping" a bit.
Maybe you have a similar issue? Perhaps you could share a photo of your setup?

[Medwar]

Most likely your objectives are fine.
This sort of behavior is very common with 3D subjects and microscope objectives.
To explain...

Rik,
it seems that you are right!
Thank you for the great explanation.

I was already using illumination that hits from all angles.
I tried shooting the other parts of the object and other objects, and what I discovered:
1) if an object is the end of the needle: uniformly lit with no close background and no surroundings that may block a part of light - there is no "rotation" or "squirming around" at all.
2) if I shoot a different part of the subjects, the "squirming around" effect depends on the surroundings - if there is a cavity in the specimen behind one side of the object - "squirming around" is likely to appear. If the object is partly "shadowed" by the relief, also likely. It also depends on the shape and relief of the surroundings!

What remains a mystery to me:
1) Why does the plane of the "squirming around" rotate, when I rotate Mitutoyo 10x objectives 90 degrees in their thread? The lighting remains the same. It brings up thought about assymetry of the lens. But: my experiment is not pure, as the objective after rotation inclines down a bit, because of the loose thread.
2) Why the same scene shot with the Mitus 5x and 20x and Nikon 10x does not show "squirming around" at all? The "surrounding parts of the specimen" that block parts of the light are the same in this case.

[Medwar]

I've had something like that happen when the camera wasn't properly aligned horizontally, meaning it was "dropping" a bit.
Maybe you have a similar issue? Perhaps you could share a photo of your setup?

Here is the photo of the setup, the lighting was a bit different in that there are total 5 lamps and I placed them all around this diffuser. Sigma 70-200 2.8 EX as TL here.

Instability of the rig or the specimen was my first thought!
But 5x and 20x showed absolute stability with no "rotation", so that is not the case.

[Medwar]

Perhaps you could share a photo of your setup?

And here is the overall view of the setup.
The assembly with Componon-S 100mm as TL.
Parts: stackshot controller and motorized focus block instead of the rail. Proxxon KT-70 coordinate table.
The ceramic floor tile under the setup is a must! Otherwise the plywood base of the rig bends together with the chipboard table, when I step on the floor near the table.

[Medwar]

Is it causing issues with stacking? If not I would not consider it a problem or worry much about what the out of focus areas are doing.

I have found that in my case with mineral specimens, the effect appears everywhere, with Mitutoyo 5x it is also present, it is just not so obvious.
And it may affect only some areas on the frame, specific parts of the relief, where one of the light sources is shadowed.
This doesn't affect normal stacking much (however I am not sure that there is no effect at all).

But it seems that this very effect kills the idea of stitching stereo-pairs of surfaces with complex relief (minerals) made with telecentric optics. Because even with telecentric optics, this effect seem to create areas with "fake parallax". For stitching stereos there shouldn't be any parallax at all. The tiles seams are extremely glitchy in stereo if there is any parallax. I have to run some more tests to be sure.

[rjlittlefield]

The trick for stitching 3D is to use a really good focus stepping mechanism and completely turn off computational alignment.

That way the weirdly shifted OOF regions will not affect the geometry of the stacks. Instead the rendered stack will have the geometry of the focus stepping mechanism -- ideally perfectly rectilinear. This method works best if you have a telecentric lens, but at 10X and above it also works well with lenses that are not telecentric, because the shallow DOF per slice allows the images to be combined without scale correction and still look good. (The Mitutoyo 5X is so close to telecentric that it will work fine.)

It looks to me like you're using a microscope focus block that may be good enough to make this work.

So, in Zerene Stacker just remove all the check marks at Options > Preferences > Alignment, and see how you get along.

--Rik

[CrispyBee]

This has been bugging me....

I'm torn between wondering whether it's a result of field curvature and/or an optical illusion due to the lower have of the image not being in focus when moving the camera forward.
It would be far easier to assess the real culprit if you had a small object in the center and a background that is evenly visible.

The sloped position of the subject also introduces an observer bias so our eyes will "follow" the focus upwards to the left and that makes it look like it's rotating.

EDIT: Yeah I actually think it's an optical illusion. Squinting and/or covering parts of the image makes it look "normal" with the focus racking back and forth. The shallow DOF and the fact that the specimen is sloped and the crystals in the front point to the side give the illusion that the perspective is shifting/moving/rotating but it doesn't.

[rjlittlefield]

This has been bugging me....

I'm torn between wondering whether it's a result of field curvature and/or an optical illusion due to the lower have of the image not being in focus when moving the camera forward.
...
Yeah I actually think it's an optical illusion.

Before committing to either of those explanations, I suggest to run a simple physical experiment.

In particular, set up one of your larger cameras with live view, looking through a nice wide lens at a closeup 3D scene. The wider the lens and the closer the scene, the better. Set the lens wide open and focus on the center of the scene. Now take a black card and punch say a 2mm hole in it, to act as a movable aperture. Hold the card in front of the lens, blocking all the light except what gets through the small hole.

Then move the card around, and watch the live view.

What you'll see is the rather surprising effect that the OOF foreground and background elements move around in sync with the hole in the card. Features that are perfectly in focus stay in one place, while OOF background moves in the same direction as the hole, and OOF foreground moves in the opposite direction. If the lens is close enough to the scene, you can even change the angle of view, for example making OOF foreground bigger but OOF background smaller, by changing the distance from the hole to the scene. All of this happens with no change to the camera, the lens, the subject, or the lighting. All that moves is the hole in the card.

I have placed an example at https://janrik.net/MiscSubj/2023/Moving ... erture.mp4 , but I really encourage you to do the experiment yourself.

What's going on with that experiment is explained in an article that I wrote almost 18 years ago, Theory of the “No-Parallax” Point in Panorama Photography. Very briefly, the apparent position of OOF features depends on exactly which light rays make it through the lens to strike the sensor. Moving the aperture selects different rays, which causes the OOF features to move around.

In cases like medwar's, there is no obvious moving aperture. However, there are asymmetries that cause less than the full lens aperture to be used for some parts of the subject, either because some light is blocked by other parts of the subject, or because light bounces off the subject in such a way that it only enters part of the lens. It is partial use of the aperture due to those asymmetries that cause OOF things to move around as focus is changed.

--Rik

[Medwar]

Hi Rik!

Your advice to turn off computational alignment completely for stereo stitching hit the bull's eye! Turning alignment off completely is really a must.
I return to my photo-hobby from time to time when I have gaps in my work and family life. I returned to this topic recently, and I would like to thank you once again.

However, I am experiencing difficulties with choosing the stitching software, none of them does good for stereos. PTGUI has the best seams quality, but creates terrible barrel distortion all over the frame, not only in intersecting areas, regardless of settings that I tried. I think I will start another topic about it with illustrations, maybe after running some more tests.

[rjlittlefield]

Thank you for the feedback.

I will be interested to read the new topic so that I can see what difficulties you are running into with stitching.

--Rik