I'm projecting 3D points with X,Y,Z model coordinates to X,Y image coordinates using a 4x4 perspective view-projection matrix. There is only one model, so it is like a MVP matrix where the M matrix is unity.
Is it possible to extract the coordinates of the position (in model coordinates) of the camera from the view-projection matrix. (i.e. the translation component of the view-matrix)?
Also, what exactly is the meaning of the Z-component in the projected image coordinates (after division by W)? I know it is between -1 and 1 for points between the near and far planes, but is it possible to deduce the distance of the point to the camera (in model coordinates) from it?