Stitching micro panoramas

[elf]

The technique I use for stitching macro panoramas is essentially the same as for landscape panoramas. The camera is rotated around the entrance pupil both vertically and horizontally. When focus stacking is added to the mix, I leave the lens (and subject) fixed and move the camera which changes the extension, focus point, and magnification. Zerene Stacker handles these changes quite well at lower magnifications with standard lens. This method doesn't work well at higher magnifications (10X and up) with microscope objectives because the objectives are designed to work at a specified extension. To cover even a modest depth of field with a 10X objective requires a large change in the extension and image quality suffers.

Rik asked in this thread what issues people were having with stitching micro panoramas and I'm curious to find a technique that works well at the micro level.

[rjlittlefield]

I'm curious to find a technique that works well at the micro level.

For micro, you've gone down the wrong path by adjusting focus via bellows draw.

Instead, just treat the camera and lens(es) as a fixed unit. Adjust focus by changing distance between the subject and the lens. When running the stacking software, turn off all alignment -- no shift, no scale. The result is that each stack will produce a result image that has exactly the same geometry as a telecentric lens. Collect any number of such stacks by just shifting the subject in X and Y. The resulting stacks will stitch up cleanly without parallax errors. If there is barrel or pincushion distortion, then depending on the stitching software you may need to correct that as a separate step before stitching. (All the stitching software I know can correct distortion on the fly when images are shot with the normal spherical geometry, but some of them cannot correct distortion on the fly for a "flat stitching" problem like this one.)

My avatar comes from a stack-and-stitch effort as described above. I forget exactly how many stacks it was -- around 16, I think. The effort was described in-progress HERE.

--Rik

[Cactusdave]

I'll lay out the problems I've had. To be honest I don't think they are necessarily problems, more revealing of my own technical deficiencies and lack of patience when acquiring images .

1. Good panoramas require more overlap between individual images than you think, good micro-panoramas require a lot more in my experience.
2. Good stacking requires good overlap between individual depth frames, equally true when stacking under the microscope.
3. A very systematic approach is necessary as during the stacking acquisition at each point it's fatally easy to forget where you are in the panorama acquisition.

1X2 = An awful lot of images and a lot of post processing before even a raw micro-panorama emerges. For example a subject that requires 100 images at X10 to cover it adequately, and 10 stacked images at each point to reveal its depth then suddenly we have 1000 images to work on, and this is not untypical.

Consequently there is a fatal urge (in my case at least) to reduce this by acquiring too few images, resulting in disaster at the stitching stage or at least obvious stitching artifacts. Precisely, of course, when the hard work has all been done and can't be undone.

The fun doesn't stop when you have your initial stitched image. These are big images, gigabyte images and they give even a reasonably powerful computer pause for thought., Cleaning up and spotting a 15000 X 10000 pixel image at the pixel level is masochist territory.

Finally the way we work with images from the microscope which inevitably have a lot of dirt spots from the slide mount and the camera sensor means we pixel peep a lot. This can lead to disappointment about stitching boundary glitches and glitches in depth of focus continuity that when looked for at monitor scale are invisible, but annoy us perfectionists terribly.

None of these things are really technical issues, they are issues of patience and desire to make the best image we can. Nothing frustrates like perfection not achieved