Reconstructing Colored Strip-Shredded Documents based on the Hungarians Algorithm

One of the common problem in forensic science and investigation science is reconstructing destroyed documents that have been strip-shredded. This work intend to design a strip matching algorithm to resemble edges of the strips, in order to reconstruct the original document. The proposed algorithm is divided into three phases. First is image-based similarity evaluation to produce a score function (which includes building “distance matrix”). The second phase is assignment phase that matches a shred border pixels of the right side to a left side of another shred (using Hungarians algorithm). The third phase is defining the sequence according to the matched strips in order to merge the shreds and reconstruct the document. The proposed work is compared with a nearest neighbor search algorithm in term of accuracy and speed. The Hungarians reassembling algorithm scores better accuracy and run time than nearest neighbor reassembling. The proposed approach scored (96.2 percent) as an average accuracy for reassembling an available online benchmark.