I believe that to do this is virtually impossible.
You would need a means of locating straight lines in an image; such algorithms exist, but they are pretty complicated.
Having that in hand, you would need to recreate the image repeatedly, with every permutation of removing lines implemented (one without line_1, one without line_2, one without line_1 and without line_2, etc., up to the number of lines that were detected).
Then each of those images would need to be passed through Optical Character Recognition.
I think you’re going to need the help of the CIA, the FBI and MI6...