I wrote a program to do this two decades ago. I shredded a black and white text document (long shreds, not the new shredders that cut into confetti), scanned all the strips, front and back, and created a lookup value for each edge based on the pixel position. The biggest problem I had was the shredder blade removed too much material. This caused a continuous stroke to be fatter on one side than the other (imagine if the descender in a lowercase "p" was on the blade's path).