Graphic explanation of why 3d images can appear "deeper" than they are ("dept exaggeration", or "vertical exaggeration" in case of planetary images taken from orbit):
Attivare Java per una costruzione interattiva (con Cinderella).
Created with Cinderella

Once you fix the positions of Near Object, Far Object and Camera, and you shoot the photo, you obtain A and B points on your picture: they represent Near and Far objects, and their position is fixed ("stored" on the picture).

When you look at the picture, visual rays which connect your eye to the object must be the same of the visual rays connected the camera "eye" to the objects (distance does not matter, as long as angles are the same).

If you are too far from the picture (try moving down point named "Eye" in above java applet), the "virtual objects" your eye sees get more distant, so your brain sees more distant objects.

At the opposite, if you are too close to the picture, virtual objects will come closer, and your brain will see closer objects.

If Far Object is the bottom of a mountain and Near Object is its top, getting out from the picture will result in "top more far from bottom illusion", i.e. "vertical exaggeration"

In 3d pictures you have two images, one per each eye; this is shown in this (quite more complex) schema:

Attivare Java per una costruzione interattiva (con Cinderella).

Excel 95 spreadsheet to calculate exact viewing distance for a stereopair, given:
- camera distance from object
- inter-cameras distance
(thanks to CosmicRocker user from this forum)