Trying to hand make a perspective accurate texture would take time, but it would also be a bit futile... the point is as we move our brains expect the perspective to cause a corresponding change of image, so of course a static image designed for one perspective is insufficient.
I am talking about encoding a texture so that it contains depth information. Actually I just came across this (hence me revisiting this post after a few weeks)
Basically, I invented a concept that had already been invented ... again .. sigh ...
Its called parallax shading.. I suppose in some ways its not dissimilar to bump maps in principle.