08/12/2017 (Sat) 16:57:02
Sure, it's just a simple equation. In the pictured image it's a mere (g.field) sum. To have multiple resulting plaintexts you would add another equation to the system.
Say A and B are the images, x and y the ciphertext and keystream (in F(2^n)), then the pictured example is just a trivial:
A = x + y
To also have B, you would add another constraint, say:
B = (x << t) + (y % 2^(n-t))
Where t is the offset for ("3/4 to the right").
Strictly speaking this would not exactly describe the pictured example, because there the displacement is not linear but 2-dimensional, but we can sidestep that by instead considering A(i) and B(i) (and x(i) and y(i)) for each "i" pixel row in the images.
If you keep adding constraints you might eventually make the system unsolvable, but for two offset images it's pretty much trivial.