Why don't you move camera and lens for the stitch? Like you would in an ordinary panoramic stitch
I am doing this now (pan subject in X and Y, move camera in Z) at 2x with the 95PN, and indeed it works reasonably well, but lighting must move with the subject in order for the stitch to go together without visible stitching artifacts. This is highly constraining, and requires very large working distances to get lighting between lens and subject. The lens also
should be telecentric to minimize distortions and light angle errors. The 95PN is not, and about half or more of my panoramas must be re-done due to noticeable lighting shifts.
Panning the camera in X and Y and subject in Z allows me to keep the lighting fixed and optimized. There is essentially just a single image to work with, and if I had a big enough sensor I could shoot it all at once, but alas 60x60mm sensors are not readily available.
Ultimately I don't know if this method will do any better than moving the subject. There is also a subject size limitation with this method, though for larger subjects I can just reduce the magnification, assuming I have suitable optics with sufficient image circle at that mag.
Why this give any inspiration
http://microsculpture.net/
Wonderful video, thank you so much for sharing that! Also it is a huge amount of work for each final image, much more than I'm willing to put in. I currently am doing 2x3 panoramas, with each tile composed of 10-12 shot stacks, for total of 60-72 images. At 3x I will need to do 4x6 panos due to overlap inefficiency, and depending on the quality of the objective I may need to do up to 20 shot stacks, so 480 total images. This is getting big, but still not the 10k+ as shown in the video.
How about a line scan lens, there are some with around 80mm image circle...?
Yes, most likely this is the best source for such optics. In fact the Rodenstock 105mm f4 Inspec.x L 3.5x that I am currently planning to use was specifically designed for 82mm line scan applications. The reason I'm looking further is that at f4 its NA is still fairly modest at 0.1, which is the same as the 95PN at its optimum f3.3. Thus in theory I won't be getting much more information than I am getting now with the 95PN.
