From what I've seen normaly the mental model is that you are
the one moving not the picture [...]
Would be interesting too look at why this is, is it conditioned
because this is how most software acts, because of scrollbars
or just because the user moves the mouse in the way they
would like to move if the image were real.
I think Pacman is part of the reason...
In an everyday situation, you would typically move the
object. Say someone puts a photo on your desk and asks
you to take a closer look at some detail. Would you bend
over to get a closer look, or would you take up the photo
and hold it closer to your eye?
My feeling is, by far the most of us would lift the photo.
Perhaps we would later put it back on the desk and
bend over it using a magnifier, but that's because of
practical difficulties with light and keeping both the
photo and magnifier steady, not because bending feels
natural.
Say you spot a strange insect walking your desk. If
you're the curious kind, you'll bend over and watch,
but that's merely because you've been taught not to
touch such things. A toddler would not hesitate to
take it up and pull it apart in front of his/her eyes.
Now, in 3D games, it's an established convention that
you move your alter ego around by rolling/pushing/
pressing in whatever direction you want to go, though
this is the opposite of what you would do with your
hands if you were really crawling in a dusty tomb.
It's not because this is the "natural" way. It's a convention
inherited from Pacman where your avatar is more like an
object you move around.
As things are, with Counterstrike, Hitman and their likes
being the typical youngster's first exposure to computers,
this little twart is probably here to stay.