X-rays and Machine Learning Unfold Historical "Letterlocked" Documents — Entirely Virtually
High-resolution X-rays and some clever computation allow letters hundreds of years old to be unfolded and read without damage.
A team of historians, computer scientists, and dental imaging specialists have published a paper detailing a method for reading folded letters — even those secured through "letterlocking" — through sealed envelopes and letterpackets, allowing fragile historical messages to be viewed without damage.
"Letterlocking was an everyday activity for centuries, across cultures, borders, and social classes, and plays an integral role in the history of secrecy systems as the missing link between physical communications security techniques from the ancient world and modern digital cryptography," the researchers claim in their paper's introduction. "While our attention is naturally drawn to a letter’s written contents, the material evidence on surviving opened letters, such as crease marks and wax seals, testifies to thousands of folding techniques used over the centuries to turn a flat sheet of paper into a secure letterpacket."
"With careful study, this evidence can be used to reverse-engineer historical letterpackets, which themselves become a key dataset for the study of historical communications security methods."
The first step, though: Being able to image the letters without taking them out of their envelopes or unfolding the complex letterpackets and risking destroying them completely. "I work in the dental school at QMUL [Queen Mary University London]," writes Dr. David Mills of his work on the paper. "I'm not a historian or an archivist or conservator, so why was I involved in scanning these letters?"
"Calcium is a metal, it absorbs X-rays. Iron and Copper also do to, and they absorb x-rays better than calcium does (simplification). This means we can see tiny traces of Fe and Cu in the Iron gall inks used on parchment and paper in history. As teeth decay, they go though a series of processes were they lose mineral content. Our scanners can see this tiny change, our department does a lot of research on this. A side effect is, our scanners are so sensitive they can be used for projects like this."
With the scans in hand, work turned to not only reassembling the data into the folded letter but unfolding it entirely virtually — using computational flattening algorithms, presented as a fully-automatic system that takes the scan in at one end and spits out a 3D reconstruction, a corresponding 2D representation of the letter in its unfolded state, and a mapping which shows how the folds relate to the unfolded letter.
The paper has been published in the journal Nature Communications under open-access terms; more information is available on the Letterlocking Twitter account.