A moving plane takes a picture of the ground. Then it flies a bit forward and takes another picture , the second image having a common part with the first (they "intersect" in a polygon shape).

Problem: given 2 such images determine the intersection polygon.

My question is , since a pixel in the image is an interpolation of a real life area, the "intersection" of the 2 images may be composed only of different (but similar) pixels.
May I be the intersection , I in image one is fairly similar but no 2 pixels have the similar value as I in image 2.

What distance can I chose for such a situation ? Can anyone point me to useful approaches to this problem ?