When reconstructing 3D features using epipolar geometry image points are used to derive the 3D geometry. If the input images are poor there may be not enough image points available to produce a mathematical solution. With only few degrees of freedom left it still might be possible for a human to match the images sufficiently.

I have a set of images as pictured below, where only two edges and their intersection point can be identified reliably. These two edges form an angle of 90° in 3D.

1. Is it possible to use edges / lines of the input images to aid in calculation of epipolar geometry?
2. If so, how many degrees of freedom would be left in this case?
3. How many DoF would be left if both images have the exact same camera parameters?