In theory you could have some sort of multi head or rasterizing system scan every bit of the tape and construct a 2D map of the magnetization. It would have to be 2D because many tape formats have multiple tracks, not even in parallel. But decoding that data would be a nightmare that made optical character recognition look like child's play.
Really? Couldn't you just 3D model the player and run the simulated tape through it? Pretty amazing if there's something a 70s mechanical device can do that a computer can't do easily.
