Does the location of a virtual image through a curved mirror depend on the observer's viewpoint?

Countless ray diagram demonstrations on the Internet simply ignore the observer when it comes to curved mirrors, and they’re simply about convenient rays (parallel to the principal axis, converging at focus, etc.), which gives me the impression that the location of the virtual image is at one place, no matter where the observer observes it. I’ve only found one diagram that specifically says the observer’s view point must be on the principal axis for the ray tracing method to work; and this paper that says, unlike planar mirrors through which the virtual image doesn’t ever change location, virtual images through curved mirrors change location depending on the observer’s viewpoint (look for section "2.3. Planar Reflectors").

So does the location of a virtual image through a curved mirror depend on the observer’s viewpoint? If that’s the case, if the viewer’s eye is not on the principal axis (which according to pretty much all demonstrations out there, is always positioned so that it goes through the center of the mirror’s surface for no apparent reason as far as I can tell), should I choose my own principal axis that goes through the viewpoint for the ray diagram method to work?