A pipeline for reconstructing cross-shredded English document

Document shreds reconstruction is of great significance in the fields of file confidentiality, anti-disclosure, and investigative science. In this paper, a complete and practical pipeline is designed to reconstruct cross-shredded English documents. The pipeline firstly classifies the shreds into several clusters based on an improved K-means algorithm to reduce clustering imbalance. Especially, a preprocessing is needed before extracting feature vector for shredded English document because of the unaligned characters. Owing to its successful performance in reconstructing the strip-shredded documents, Hungarian algorithm is applied into the permutation for the cross-shredded shreds in the same row. Eventually the location of the connective horizontal paper slips are arranged by considering the complementary relationship of edge vectors between two neighboring shreds. Reconstruction experiment results indicate that the designed pipeline can acquire high precision and efficiency.