Features based reconstruction technique for torn documents
Marwan D. Saleh, N. D. Salih, Wan-Noorshahida Mohd-Isa · IET conference proceedings. · 2023
The reconstruction of torn documents that hold sensitive information or crucial historical data is of great significance in many domains. An efficient technique for hand-torn document restoration is presented in this paper. Pre-processing, feature extraction, and matching three main phases of this technique. The reconstruction is carried out by matching pairwise the feature vectors of the fragments. Using a database consists of 120 torn pages taken from books, newspapers and magazines, the performance of this technique has been measured using simple but effective criteria called error rate. The algorithm was able to accurately reconstruct 91% of the documents, with the remaining being partially reconstructed.