The context might be an old photograph of (say) soldiers in a trench or a picnic in a forest, but without a horizon or any vanishing points.
My conjecture is that 4 known points in world space mapping to 4 points in projected image space are enough, with the constraint that the points form a solid object and a quadrilateral respectively.
The aim is eventually to find the location of the camera and its field of view, to reproduce the scene in software.
Am I right, and if so how to perform the calculation?