projective transforms hidden under affine_transform

[VolkerH]

This is not a bug report but rather a comment about nomenclature / usability.
I was looking for projective transforms in scipy.ndimage and initially didn't find them, so I thought I would have to go with geometric_transform.
Only when looking at the documentation for affine_transform later did I notice that affine_transform also supports homogeneous coordinates and thus actually supports transformations that are more general than affine. I feel that there should be something like a projective_transform function that defaults to using homogeneous coordinates.

[ilayn]

If I understand you correctly:

From the affine transform y=A@x + b, if you don't include the offset which is b and 0. by default you have projective transform A@x. Hence it already defaults to homogeneous coords.

[VolkerH]

My point is that a homography is more general, as it doesn't necessarily keep parallel lines parallel (which an affine transformation does).
Edit (got interrupted while writing this):

So when looking for available transformations by name it isn't obvious that you can actually use affine_transform for projective transformations, this only becomes clear when reading the detailed documentation that mentions the possibility to use homogeneous coordinates.