A gigantic puzzle lies stored in the archives belonging to the Agency of the Federal Commissioner for the Documents of the State Security Service of the former German Democratic Republic (BStU). There are about 15,000 sacks full of shredded documents from the former Ministry of State Security, the internal and foreign intelligence service of the GDR. Each of the estimated 45 million DIN-A4 pages were sliced into between eight and 30 pieces when the East German dictatorship disintegrated after the fall of the Wall in 1989. Only a fraction of these documents have been successfully reconstructed so far, because the paper puzzle is too complex. It would take 30 people 600 to 800 years to put the estimated 600 million pieces of paper back together again. But scientists at Berlin’s Fraunhofer Institute for Production Systems and Design Technology (IPK) have developed a computer-aided process to analyse the mutilated pieces of the historically valuable documents automatically and speed up the recovery effort. The scientists were commissioned by the German Bundestag and the BStU to construct a so-called ePuzzler. The application is currently in the pilot phase, but initial results are already very promising.
Before the digital reconstruction process can begin, each shred of paper has to be scanned, but this has been made simple thanks to a newly developed scanner. The snippets are drawn into the scanner and scanned on both sides without having to be laminated or treated in any other way beforehand. The digital images are then analyzed, sorted into categories and stored in a database. The Fraunhofer system uses a great diversity of features in the search for matching pieces. They range from the shape of the snippets to the colour and individual motifs on the paper. When the contours of two puzzle pieces coincide, the computer fits them together to form a larger piece and continues this process until finally a whole document is complete. Then the process starts again with the next document.
It’s anticipated that the pilot project will be concluded in the second half of 2010. It will then be decided whether the system will be used to analyze all of the decimated documents.
