May 2025 – Digital Palaeography

In manuscript studies, scholars often encounter texts that are barely legible from centuries of damage, reuse as in the case of palimpsests, faded ink, or chemical reactions. Multispectral Imaging (MSI) has emerged as a way to recover such hidden texts, capturing images of manuscripts under different wavelengths of light to reveal what the naked eye can no longer perceive.

While MSI is a successful and useful technique for text recovery, the tools required to process MSI images are often intimidating. Many existing applications assume a high degree of technical knowledge, from a chemical literacy to skills in navigating dense software menus. This steep learning curve can be a barrier for students and researchers who are new to the field.

A Tool for Everyone

Our goal was to develop a lightweight app with a simple, intuitive interface that enables users—especially those without a technical background—to explore MSI data without coding. We envisioned Proteus not just as a tool, but also as a guide: something that helps users navigate MSI post-processing with ease, without becoming imaging experts, by giving simple instructions on what images or features should be used for certain scans.

Instead of offering dozens of rarely-used features, Proteus focuses on a few impactful tools, delivered through a simple interface. We designed the interface so that each function is explained via pop-up tips, recommends effective image combinations for different tasks, performs the heavy computations silently in the background, and is supported by our MSI tutorials for a better understanding of the whole process. Our aim was to allow scholars to focus on interpreting texts, not on configuring software.

Simple and Sustainable Design

Proteus is built in Python, using libraries like OpenCV, NumPy, Pillow, and tkinter. This combination allows us to offer real-time image processing inside a platform-compatible graphical user interface. The GUI, powered by tkinter, makes image enhancement accessible through point-and-click interactions— which comes in handy for users with humanities backgrounds.

Users can load certain image stacks taken using multispectral imaging systems into the app, choose which feature to use, and select bands for further processing.

Among Proteus’s core features are:

– Pseudo-color images, which enhance contrast by combining spectral bands as RGB channels.

– Principal Component Analysis (PCA), which mathematically isolates features that reveal faded or overwritten text.

– Power, which intensifies each pixel in the image to get a better result, especially on faded pages.

These enhancements are powered by matrix operations (via NumPy) and OpenCV functions for tasks like histogram equalization, resizing, and denoising. Proteus also follows standardized naming conventions for exported images, and delivers a text file with the metadata, including a record of used features. This makes it easy to manage results across projects.

The app’s modular code design ensures future extensibility—whether for supporting more spectral formats, integrating machine learning, or adding new filtering methods.

The Current Version of Proteus

The first working versions of Proteus have already been actively used the app in internal projects and tasks. Following evaluations with Dr. Anna Dorofeeva and Dr. Alexander Zawacki, and with the assistance of our MSI team, we’ve gathered valuable feedback—pinpointing both bugs and areas for improvement.

This feedback is now shaping Proteus v3.0, which will feature improved performance, bug fixes, and new features inspired by user experience. We’re also excited about early experiments using Proteus on non-MSI images, where enhancements like PCA and contrast filtering have revealed details even in ordinary scans.

What Next?

Proteus is built in accordance with the principle that digital manuscript tools should be clear, simple, and powerful. We want to minimise the time spent on post processing and allow researchers to make the best use of their time.

We intend for Proteus to become a standard teaching and research tool in digital humanities programs, especially in MSI training units. By simplifying complex workflows, it will lower the barrier for students and scholars – even those with limited technical skills – to engage with multispectral imaging.

As the field of manuscript studies embraces new technologies, we believe tools like Proteus can lead the way—quietly doing the heavy work, while putting discovery back in the hands of the user.

References

Conjeti, Sailesh, Stefanie Demirci, and Vincent Christlein. Image retrieval in big image data. Elsevier, 2023.

DigiTRiP-Redaktion (17. Februar 2025). Proteus – Handschriftengeheimnisse aufdecken. DigiTRiP. Abgerufen am 26. Mai 2025 von https://doi.org/10.58079/13bto

Knox, Keith. “Hoku Software .” Rochester Institute of Technology, 11 Sept. 2017.

Leslie’s Blog. “Gradual Gaussian Blur Using Opencv.” Leslie’s Blog, 5 Aug. 2020, leslietj.github.io/2020/08/05/Gradual-Gaussian-Blur-Using-OpenCV/.

Roepke, Brian. “The Magic of Principal Component Analysis through Image Compression.” Lambda with Text - Wide, Data Knows All, LLC, 22 Mar. 2024, dataknowsall.com/blog/imagepca.html.

Team, PlantCV Development. “Pseudocolor.” PlantCV, plantcv.readthedocs.io/en/latest/visualize_pseudocolor/

One Year Anniversary of Project eManuSkript: Part One

In April 2024, a group of scholars and students, funded by the Stiftung Innovation in der Hochschullehre, came together at the University of Göttingen’s Institute for Digital Humanities with an ambitious goal: to reimagine how medieval manuscripts can be taught, studied, and experienced in the digital age. Led by Dr. Anna Dorofeeva, Project eManuSkript has grown since then into a vibrant, interdisciplinary project that continues to evolve—shaped by its users, enriched by its team, and rooted in a shared commitment to accessibility and collaboration.

The Vision Behind eManuSkript

The idea was simple but important: manuscript studies, often remain difficult to access, especially for beginners or those outside of specialized institutions. Scholars might face barriers due to geographic distance, limited physical access to manuscripts, or a lack of training in palaeography or codicology.

Our goal was to build tools that would meet learners and researchers where they are—whether in the classroom, at an archive, or online. We wanted to create a digital platform that was not only technically sound but also inviting, accessible, flexible, and pedagogically grounded. The user experience and the needs of students in manuscript studies have driven the development of the project.

• • •

What We’ve Built

Over the past year, the eManuSkript team has been developing a suite of interconnected tools designed to address a range of issues associated with teaching and learning medieval manuscripts. Each one tackles a different challenge, and together they form a single interface.

Mergen is our AI-based layout segmentation app. It detects key manuscript elements—both text and paratext—with exceptional speed and accuracy, outperforming other existing segmentation tools and datasets. By using Mergen, the user can build a research corpus of specific manuscript elements. Using a three-layer segmentation method, our team annotated a significant corpus of raw manuscript pages to make sure Mergen reaches its potential in the shortest time. We still continue to update Mergen’s database regularly for even better results. ㅤㅤㅤㅤㅤㅤㅤ

Proteus simplifies a complex task: working on texts that are hard-to-read, whether due to chemical damage or purposeful erasure, such as in palimpsests, by making use of multispectral imaging technology. With Proteus, users can generate enhanced images—for example, using pseudo-colour or PCA—that reveal hidden or faded script. Proteus can even enhance high resolution visible-light images, using power or contrasts. Currently, we are finetuning Proteus V3.0.

Seshat is an annotation-and-measurement tool that allows users to analyse historical scripts. Users upload the manuscript they want to study using a IIIF link. They can then trace letterforms, calculate letter size, generate statistical data such as stroke angle, and compare different scribal hands. The resulting data, such as tracing, measurements and image embeddings, can be exported to be analysed further. We are also working on a script recognition model to add a precise, compact model of Latin scripts to our web app. A downloadable version of Seshat will soon include trained models for text recognition.

Ceres &Libera are two connected applications for exploring medieval bookbinding. Ceres, allows users to generate customised templates to visualise medieval manuscript bookbindings. For this app, our team’s plan to document the spine and sewing structures of a medieval book have been supported by expert input from conservators at the Göttingen State and University Library. The second app, Libera, is an interactive application where users can build their own codex from bookblock to cover. With Libera, users will be able to follow simplified tutorials as they build their codex and can build visualizations with specialized terminology and professionally illustrated components.

Supporting these tools is a rich range of interactive tutorials—including images, videos and interactive games—as well as a historical timeline and canonical script alphabets with ductus practice options. Accessibility is built into every layer: visual alternatives for those with impairments, multimodal learning options, and a user-friendly design that reduces the barriers to working with manuscripts.

These applications reflect our experimental approach to manuscript studies. Despite the growing number of digitised collections, access is still limited by the high level of specialist knowledge required. Through regular team meetings, in-class experimentation, and constant dialogue between developers, scholars, and students, we’ve shaped tools that genuinely respond to user needs—pushing boundaries while remaining grounded in traditional scholarship.

Together, these five tools create a full digital environment for investigating the medieval manuscript book—from physical structure and visual layout to textual content and palaeographic detail. By April 2026, all applications will be available through our website, and our datasets and backend code will be archived openly via GitHub and Zenodo. This suite of tools and tutorials will be complemented by an extensive glossary and Zotero library. We hope this will empower others to build upon our work and bring manuscript studies fully into the digital age.

• • •

People and Collaboration at the Heart

What truly powers Project eManuSkript is its team. Beginning as a group of ten, with backgrounds in history, digital humanities, computer science, German studies, English literature, and more, our team has changed over time, but what has remained consistent has been the interest the project has generated — from students to guest researchers, all enriching our ideas with fruitful support and feedback. Our collaboration is intentionally non-hierarchical and inclusive. We believe that the best solutions emerge from diverse perspectives and shared responsibility.

Team members contribute in many ways: designing user interfaces, labelling datasets, testing OCR models, writing documentation, or organizing outreach events. Each of us brings a unique angle, and we encourage constant feedback—especially from our fellow students at the Institute of Digital Humanities, who are both co-creators and core users of the tools.

• • •

Challenges and Lessons Learned

Of course, no digital project is without hurdles. Training machine learning models on historical scripts has meant confronting the limits of existing datasets and annotation tools. Designing interfaces for a multilingual, cross-disciplinary audience has forced us to think carefully about clarity and adaptability. And balancing research with usability remains an ongoing effort.

But these challenges have been productive. They’ve taught us to take a step back and recalculate, document thoroughly, and centre accessibility at every stage. We’ve also learned the value of working in public: sharing early prototypes, inviting critique, and adapting based on what users actually need.

• • •

What’s Next?

As we move into our second year, we’re expanding in different directions. This year, we’re refining our tools with the help of classroom testing, user studies, and more targeted community engagement. We are hoping that, by releasing our website for Project eManuSkript, we will encourage a larger public to explore our current work, both to offer constructive criticism and feedback on what we have accomplished so far and to measure their accessibility and practical usability for scholars and students from all around the world.

We are continuously working to make the best out of each app and expand the limits of our tutorials by going beyond the traditional approach. Going forward, we will continue to generate new ideas, to maintain outputs that are open, free, and adaptable, and, ultimately, to develop the best technologies we can for digital manuscript studies.

•

A Thank You—and an Invitation

Dear Reader,

On this first anniversary, we want to thank everyone who has contributed to eManuSkript: our funders, testers, advisors, and especially the students who’ve helped shape the tools with their questions and insights.

This project is still growing, and we hope it continues to spark curiosity, empower learners, and connect people to the world of medieval manuscripts in new and meaningful ways.

Whether you’re a student, a teacher, a researcher, or just a manuscript enthusiast —welcome. You’re invited to explore, learn, and build with us.
Best wishes,

– Project eManuSkript Team

Month: May 2025

Proteus: A Tool to Simplify Post Processing of Multispectral Images

One Year of eManuSkript: A Digital Journey into Manuscript Studies