[mods: sure this is long and off-topic, but I felt the need to correct some inaccurate statements]
Diagnosis correct. Remedy not.
Diagnosis is correct. A camera (a normal one, see below) projects the 3-d world onto a plane. If you rotate the camera, you change the plane, and the geometrical transformations do not match. Yet in all cases, straight lines in the world are mapped to straight lines on film. Won't go into more details, would require preparing figures, and even so... The laws of perspective are known since Alberti and a few other worked them out in the 15th century.
Remedy (move camera with film in same plane) is not. Demonstrated with a counter-example. Imagine there is a house on the right of the field of view of the first picture; far enough to the right that you see its left-side wall. Now move your camera to the right, keeping the film plane parallel, until the house is in left of the viewfinder. Now you don't any more see the left wall of the house. How could you possibly stitch these two images together?
Go back to the camera in a fixed place, "just rotating" between frames. Take
overlapping pictures (1/3 of a frame
overlap recommended). Feed them to a piece of software called a pano stitcher. The software will "undo" the projection from the (two or more) film planes onto a sphere (as above, don't ask for more details), and, from the overlap areas, figure out how the images should be positioned. It will also figure out more subtle things, like the distortion of the lens; I mean true distortion, as in pincushion of barrel, not the "distortion" of wideangle lenses. Once the images match on the sphere, the software will re-project them onto...
- A plane, effectively providing the picture that you would have obtained with a view camera. You can choose that plane, giving the equivalent of shifts/tilts. Or...
- A cylinder, effectively providing the picture that you could have taken with a Horizon or Noblex camera.
- Several other options.
And, what is meant by "just rotating" between frames? That the camera stays at the same place. But which part of the camera? If you rotate the tripod head, the front lens will rotate
and move. So we have to a small degree the unwanted effect described above in the example of taking a pano of a house by parallel shift. Over the internet, so many
experts state (parrot) that the nodal point of the lens is what must be kept fixed, while truly it is the entrance pupil. What is the entrance pupil? stare into the front of the lens, with the diaphragm partly closed; the entrance pupil is where the diaphragm
appears to be, as seen through the front group(s) of lenses.
This issue of what part of the camera must remain at a fixed position is significant only if there are nearby foreground objects.
Pano software (just example that I have used): PTGUI (shareware), Hugin (freeware).
Example Piazza dell'Anfiteatro, Lucca, Italy
View attachment 327098