This is pretty cool. Back in the nineties when I worked at Canopus, a very clever Japanese engineer named Tom wrote a software defined video routine that would read the raw ouput of a sampled composite video stream and decode it to an RGB stream at a higher quality than what the hardware could do. This was a nice competitive advantage over other cards which used the same chip for composite video recording.