Evaluation and reconstruction of strip-shredded text documents

Abstract

In my master's thesis I elaborate on strip-shredded text document reconstruction. Contrary to conventional document reconstruction -- which uses color or shape information of images -- text document reconstruction has not been researched very well. Nowadays it is common to destroy paper documents by shredding them, i.e.<br />producing paper strips. This work tries to find ways to undo the process. First and foremost I describe the problem formally. Next I define a way to evaluate problem instances. A set of improvement strategies are introduced which help the evaluation process. Defined construction heuristics yield good results in reasonable amount of time.<br />Then optimization algorithms try to find a good arrangement of the strips, ideally the correct one. A demo application simulates the shredding process of a sample page. Then this page is reconstructed using the above mentioned evaluation techniques and several optimization techniques like multistart variable neighborhood search, simulated annealing and iterated local search. Extensive tests were run with a 60 instance test set. The implemented application reconstructed more than half of the problem instances correctly and is also able to reconstruct several pages at once.