Clipping
The clipping process removes the objects in the world that the camera
does not see. The objects retained are those contained within a
frustum-shaped view volume. Usually this volume is
mapped to a canonical view volume.
Projection Normalization
In order for clipping to take place, the transformation from eye space
to clip space must retain depth information. An ordinary perspective
projection onto the view plane would lose the object depths. Instead,
a 3D perspective projection is used. This is defined in such a way
that a 3D perspective projection followed by a simple orthographic
projection onto the view plane produces the same result as a
perspective projection of the original world. The 3D perspective
projection preserves the relative depth order of objects. This allows
clipping to take place after the perspective calculations have been
essentially completed. Projection normalization simplifies hidden
surface removal. (why?)
