Reconstruction of Shredded Figures based on Grayscale Matrix

With rapid advances in information technology, image-processing methodologies have assumed an increasingly central role in forensic analysis, the restoration of historical documents, and the extraction of intelligence. In forensic applications, such methodologies facilitate the examination of evidentiary materials; in cultural-heritage contexts, they support the recovery of damaged documents. Manual reconstruction of shredded documents is laborious and time-consuming, and extant automated approaches continue to encounter difficulties in edge matching, preservation of semantic consistency, and multi-modal data fusion. This paper presents a critical review of the strengths and limitations of current shredded-document reconstruction algorithms and proposes methods specifically adapted to mechanically shredded pages. In simulated experiments, edge features are extracted from real images via clustering; a combined metric that integrates a sum-of-differences measure with correlation-coefficient-based matching is employed to assess boundary similarity and to effect both horizontal and vertical reassembly of mechanically strip-cut fragments. Experimental results indicate that, for single-direction (vertical) strips in both Chinese and English, the conjoint use of binarization and grayscale-matrix edge analysis together with sum-of-differences and correlation matching permits reliable left-to-right recursive reconstruction, achieving complete restoration on the evaluated dataset. For cross-cut fragments, K-means clustering of row features combined with an iterative correlation-threshold procedure reduces the requirement for manual intervention by more than 40% while maintaining reconstruction accuracy. For double-sided prints, comparison of front-and-back correlation coefficients enables correct two-sided reconstruction. Furthermore, embedding edge features in a shared representation using a deep metric-learning model trained with a contrastive loss function increases single-page reconstruction accuracy to above 90%. Notable limitations persist: the constrained processing capacity of typical hardware impedes accurate modeling of gray-level transitions at true joins. Future research will simulate transition gray-value variations, incorporate edge irregularities (for example, bevels and torn edges), and extend the metric evaluation to multi-page stitching to enhance robustness and generalizability.