Abstract
Medieval paper, a handmade product, is made with a mould which leaves an indelible imprint on the sheet of paper. This imprint includes chain lines, laid lines and watermarks which are often visible on the sheet. Extracting these features allows the identification of paper stock and gives information about chronology, localisation and movement of books and people. Most computational work for feature extraction of paper analysis has so far focused on radiography or transmitted light images. While these imaging methods provide clear visualisation for the features of interest, they are expensive and time consuming in their acquisition and not feasible for smaller institutions. However, reflected light images of medieval paper manuscripts are abundant and possibly cheaper in their acquisition. In this paper, we propose algorithms to detect and extract the laid and chain lines from reflected light images. We tackle the main drawback of reflected light images, that is, the low contrast attenuation of lines and intensity jumps due to noise and degradation, by employing the spectral total variation decomposition and develop methods for subsequent line extraction. Our results clearly demonstrate the feasibility of using reflected light images in paper analysis. This work enables the feature extraction for paper manuscripts that have otherwise not been analysed due to a lack of appropriate images. We also open the door for paper stock identification at scale.
Introduction
Every sheet of medieval paper contains a hidden "fingerprint" left by the wire mould used to create it. This includes chain lines (vertical supports), laid lines (densely packed horizontal wires), and watermarks. Analyzing these features allows historians to group manuscripts, track the movement of people, and date documents with high precision.
Traditionally, capturing these features requires radiography or transmitted light (shining light through the paper). These methods are expensive, time-consuming, and often obscured by "show-through" text from the reverse side. While libraries have millions of standard reflected light images (regular photos), these are notoriously difficult to analyze because the paper's texture is faint, low-contrast, and often buried under dark ink, dirt, or physical damage.
Methodology
The authors propose a multi-step pipeline that treats image analysis as a scale-dependent decomposition problem. The core innovation is the use of Spectral Total Variation (TV) Decomposition. Unlike standard Fourier transforms that decompose images by frequency, Spectral TV decomposes an image based on the contrast and size of its structures. This allows researchers to "peel away" high-contrast features, such as dark ink and surface degradation, to reveal the low-contrast, granular structures of the paper lines.
Spectral TV vs. Fourier Decomposition
A critical advantage of Spectral TV is its ability to isolate features without the "ringing" artifacts common in Fourier-based filtering. By decomposing the image into different scale bands, the algorithm can isolate the specific textures associated with the paper mould while ignoring the larger, darker shapes of the written text.
Feature-Specific Extraction
Following the cleaning process, the team employs specialized mathematical tools for different features:
- Chain Lines: The team uses Fourier filtering and projection to identify the periodic peaks corresponding to the vertical wires.
- Laid Lines: Because laid lines are much denser and fainter, the team employs the Radon transform, which is excellent at detecting linear features at specific angles, even when they are nearly invisible to the naked eye.
Key Results and Insights
The study utilized high-resolution digital images from The Parker Library at Corpus Christi College, Cambridge, testing the algorithms on manuscripts such as MS 081 (Homer’s Iliad and Odyssey) and MS 171A (Scotichronicon). These images represent real-world challenges, featuring heavy ink, discolouration, and paper folds.
- Simulating Transmitted Light: Spectral TV decomposition essentially "simulates" a transmitted light image from a reflected one by computationally removing occluding ink.
- Robustness: The paper shows that even when a manuscript is damaged or folded, the underlying wire imprints remain mathematically detectable. Switching to an anisotropic (direction-sensitive) mathematical model significantly improved the detection of laid lines in the lowest quality images.
- Successful Measurement: The algorithms successfully measured chain line distances (e.g., identifying average distances of 2.63cm and 3.93cm in different stocks) and laid line densities (ranging from 8 to 10 lines per cm), even in highly obstructed cases.
Conclusion
By providing a method that works on standard digital photos, the authors have opened the door for paper stock identification at a massive scale. While the current system is semi-automated—requiring a human to select image patches and tune certain parameters—future work will focus on full automation. This would involve using machine learning to create a "hands-off" pipeline capable of processing entire digital libraries to map the history of medieval paper production automatically.
Summary updated