European Commission logo
italiano italiano
CORDIS - Risultati della ricerca dell’UE

Non-destructive image-based manuscript analysis system

Final Report Summary - NOESIS (NOn-dEStructive Image-based manuscript analysis System)

One of the tasks facing historians and conservationists is the authentication or dating of medieval manuscripts. This task is usually possible through the study of manuscripts and the recovery of historical information such as the year the manuscript was written or cross-referencing of the facts described in the manuscripts. However, often researchers are not certain of the concurrency of the writings on manuscripts, as some writing is added at a later date. In addition, often information about the date or place a manuscript was written is not available.

In order to extract more information and determine whether certain writings are by the same scriber ,researchers often resort to the study of the type of scripting found on manuscripts. In other cases researchers compare text from different manuscripts in order to establish whether they are of the same era. To successfully address this problem, scholars are in need of scientific information, such as the type of ink used on manuscripts, that can be reliably used in the historical examination of works of art. The availability of such information would allow researchers to determine whether the writings on the same or different manuscripts are concurrent.

Most existing methods for the analysis of the material used in works of art such as manuscripts are based on destructive testing techniques that require the physical sampling of data. However, such methods cannot be used widely due to their destructive nature and the historical value of the artefacts. Computer vision techniques can be used as alternative or supplementary diagnostic methods by computing models and interpreting the visual properties of the material used such as inks.

The aim of the NOESIS project was to produce non-destructive non-invasive image-based processing techniques supplemented with historical information to aid historical analysis and examination of five significant Mediterranean collections of manuscripts hosted by libraries and museums in Egypt, Jordan, Lebanon, Cyprus, and Israel. Manuscript pages were captured under varying illumination conditions in order to model the ink appearance of segmented text and the texture of the substrate. Models of text appearance and substrate texture were formed through unsupervised multimodal clustering of descriptors and the derived probability density functions. Similarity measures were built around the distribution of cluster labels and their proportions.

The computational models augmented with historical information related to manufacture date and origin of the manuscripts and information on the scrivener formed a tool which:
1. allowed for in situ non-destructive non-invasive image-based analysis of manuscripts;
2. enabled the grouping and comparative study of manuscript ink and substrate;
3. aided authentication and dating of manuscripts from the Mediterranean region by verifying the homogeneity of their inks and substrate;
4. enabled the isolation, discrimination and analysis of faded or overlapping writings such as palimpsests;
5. facilitated the historical cross-referencing of manuscripts found in the Mediterranean region.

The system was hosted on an online database accessible via the Internet. Historical information discovered during the project was included on the database to aid future researchers. Finally, the suitability of the system was tested by using the profiling techniques for the examination of watercolours.