Restoring Harmony: The Digital Image Archive of Medieval Music

Restoring Harmony:

The Digital Image Archive of Medieval Music

Digitised manuscripts bring immense new research possibilities for scholars. One anecdote tells of a scholar who wanted to argue that a clutch of manuscripts he was studying actually came from the same source. He examined, in microscopic detail, digitised version of these manuscripts, and discovered that the hair follicles on the animal skin that formed the parchments were aligned in a similar way on each manuscript. Connecting this to other facts he had ascertained from examining the evidence, the scholar concluded that the manuscripts had a common heritage. Another project that has recently received media attention (through a series of documentaries concerning Vesuvius and Herculaneum) is the digital restoration of the Herculaneum fragments of parchment: this draws on study and techniques that have been evolved both in the USA and Oxford, and used for a number of years by the Centre for the Study of Ancient Documents in Oxford.

To allow musicologists and other related academics the opportunity to exploit these benefits of digitisation, scholars from Royal Holloway College and the University of Oxford have initiated the Digital Image Archive of Medieval Music (DIAMM). Funded originally by the Humanities Research Board, and subsequently by its successor, the Arts and Humanities Research Board, the project aims to digitise all fragments of pre-reformation scores in the UK and Europe, a corpus that has been almost impossible for scholars and musicians to access and study. Project manager Dr Julia Craig-McFeely has been at the centre of this digitisation process, using high-resolution digital camera technology to capture the images and then place them on the DIAMM website. This case study describes the creation of the DIAMM archive, from the process of digitisation (creating a digital version from the original physical object) to the delivery to end-users and investigates how the resource is being developed to suit the needs of researchers.

Digitisation

All digitisation projects are multi-faceted - issues of copyright, preservation and funding mean that there is much more to deal with than simply placing historical objects on a scanner and sitting back to wait for the finished results. Nevertheless, the actual process of digitisation, of creating the digital representation of the source, is of vital importance, as a consultation paper the team is helping produce emphasises. Without due care, the final product can be of an insufficient quality, and therefore unsuitable for scholarly use. The DIAMM team has taken the utmost care over its digitisation, believing that digitisation should entail the creation of definitive electronic versions of the manuscripts where possible. Access to rare manuscripts for the purposes of photography is often heavily restricted. The team therefore aims to create images of sufficient quality and exactitude to the original that they can reliably be used within the academic environment as long-term surrogates for the original.

Fulfilling this task has involved not the only the purchase of expensive technology, but the development of expert techniques to execute the digitisation. Such an outlay may not be a pre-requisite for all digitisation projects (indeed many projects look to outsource this work to organisations such as the Higher Education Digitisation Service), but the DIAMM project gives an excellent indication of the high quality technology needed when digitising materials that can be in an exceptionally fragile state. The core equipment used in the DIAMM project is a Fuji Pro single-lens refled camera, a PhaseOne digital scanning back, two daylight-balanced high-frequency cold lamps, two Ultra-violet lighting heads, a copy stand for mounting the camera, two Macintosh laptop computers and several hard disc drives for short- and medium-term storage and backup.

Beyond the technical quality of this equipment, speed and portability are vital attributes. All this equipment has to be carried to archives across Britain, as well as on the continent. The computers need enough hard disk space to cope with the size of the images. Each digitised manuscript in the DIAMM project (created in the tagged image file, or tiff, format, at a resolution of 600 dots per inch, as close to 7000 x 7000 pixels as possible) takes an average of 100MB in memory. Buying a computer with low processing speed would significantly impede the task of viewing and checking the images on screen and hinder the process of uploading images to the storage servers, and saving backup copies to CD-ROM.

Once Dr Craig-McFeely has her mobile studio installed in an archive and the relevant manuscript placed beneath the camera and the lights, there begins a threefold process of digitisation. Firstly, she must set, as both digital and analogue photographers would do, several variables. This includes defining the exposure, checking the focus of camera, setting the gray balance and colour calibrating all the equipment in the chain. If the exposure is set inappropriately, then the digitised object can be too bright or dark. A misgauged focus will mean that the digital manuscript will be blurred at a high magnification; poor colour calibration or unbalanced grays will mean a final product with inaccurate tones and hues. While it is impossible to create an exact replica of any physical object, with practice one can create a digital object that is very similar (even at very high magnification) to the original manuscript. Dr Craig-McFeely has refined these techniques after much practice, and is now creating digital images that surpass standards applied by many professional institutions.

When these variables have been set, the process of scanning can proceed. With such a high-resolution camera, this can take up to six minutes, and during this time Dr Craig-McFeely must take care not to make any sudden movements; any vibration transmitted through the floor can add small distortions to the image, and it is therefore not possible to use the camera where people may be walking on nearby floors. During the period of scanning, the documentation relevant to the manuscript and the scanning process is entered, on a separate computer, into a relational database holding the metadata on all the digitised manuscripts. This includes basic information usable for searching the database (the location of the manuscript, its title and contents, the rights-holder etc.) and a further slew of details related to the process of scanning (the light conditions, the various settings used, the distance between camera and object).

After the scanning is completed, quality control procedures are carried out. This is a vital process - without thorough checking of the digitised image, small errors can be missed. For example, there can be gaps or blurred bars in the image where vibrations caused the camera to shift, irregularities in tone if there were a change in the light source (sunlight appearing through a window, perhaps) during digitisation, sensors dropping out in the scanning back, or the electric current varying and causing the lighting to introduce flaws in the image. Only once Dr Craig-McFeely has examined and validated the entire image can it be accepted as a bona fide digitised copy, acceptable for depositing in the DIAMM archive.

Delivery

In an ideal world, the DIAMM team would simply upload the finished copies and their documentation to their website. However, not only does the size of these images currently make it impossible to access them across the internet, but also the widespread abuse of copyright by internet users precludes the possibility of freely distributing the digitised manuscripts: in general, libraries are wary of allowing free public access to some of their most unique resources. To be able to present the manuscripts on their website, the DIAMM team settled on a password system, which allows only registered users to access the images (although all visitors to the site can freely view thumbnails of the manuscripts). Access to the manuscripts is free, and granted to anyone with a genuine desire to study the sources; their interest is passed on to the relevant copyright holders, and they sign an access agreement which reminds them of the laws of copyright in relation to images, and also ensures that the owners of the documents will receive proper credit if any academic study is made using their sources. As the majority of the applicants are academics or other relevant researchers, the decision to allow them access to the manuscripts is generally straightforward.

As well as utilising their own website, the DIAMM is working to produce a mirror site on the Performing Arts Data Service, and will also make the searchable database of information about the sources available online as soon as possible. Documents will continue to be added to the PADS website on a timetabled basis, following practices established by the Bodleian Library regarding the release, presentation and use of authorized images. Perhaps more importantly, archiving the digitised manuscripts at another location is a crucial tool for preservation. Although the Hierarchical File Server currently used by DIAMM has some of the most comprehensive backup systems available, should the data be corrupted or affected in any way - possible as a result of the transfer process - the team can refer back to a separate copy held at the secondary storage site.

Advantages for research

During scanning, Dr Craig-McFeely is constantly reminded of the relative weakness of the human eye. At normal resolution, the naked eye is not good enough to judge the best focus or colour-scale and so to get a high-quality image, Dr Craig-McFeely must exploit the ability of the computer to detect minuscule variations in colour and focus. It is this same weakness of the naked eye, and the strength of the computer, that allows researchers to make new discoveries via the digitised manuscripts. Marks, notes, and symbols invisible to the naked eye can be revealed by enlarging these high-resolution images.

As well as digitising the medieval manuscripts, the DIAMM team is exploiting this advanced technology by embarking on a course of 'virtual restoration'. This involves manipulating the high-resolution digital images (using the graphics software PhotoShop) to enhance parts of the manuscripts invisible to the naked eye. Ink on several of the manuscripts (especially those written on paper rather than parchment) has faded, so the scholar investigating the document in the archives can make out only traces of the score. In other instances, the ink has been scraped or rubbed away: it was common medieval practice to scrape the surface of a used piece of parchment so it could be reused, often for an entirely different purpose. In some cases, therefore, the score is only faintly visible under the newer text.

Advantages for research - continued

The two images below are examples of a digitised manuscript before and after restoration. Figure 1 shows the manuscript as it was found forming a wrapping for another set of unrelated manuscripts, held at the Shakespeare Birthplace Trust at Stratford-upon-Avon. Parchments, once they had outlived their original purpose, were often used as wrapping as their durable nature made them useful for providing protection for more fragile papers. Nevertheless, using a manuscript as an outside cover is more than likely going to lead to its degradation. In this case, water damage, abrasion, over-writing and, in particular, exposure to strong light caused a drastic deterioration in the original manuscript, resulting in the image that appears in Figure 1.

Figure 1 - A manuscript, from the Shakespeare Birthplace Trust Stratford, before restoration

Figure 2 - A manuscript, from the Shakespeare Birthplace Trust Stratford, after restoration

But by using the restoration process, the DIAMM team were able to produce Figure 2, which is clear enough to indicate most of the original score. The effect was achieved by very selective filtering of the colours present in the manuscript. By filtering out the tawny colours associated with the image before restoration, and then emphasising the creamy white background and the black inks of the notes as they became more visible, the team was able to produce a 'new' version of the manuscript, with a readable version of the music. This new version, Dr Craig-McFeely stressed, was not a representation of the image 'as it once appeared' before its use as wrapping, but rather an edited version of Figure 1.

The DIAMM website will eventually feature images that represent the manuscripts before and after restoration, although there is no pressure on scholars to accept the restored image. While most scholars prefer to let others do the job of restoration, those who obtain the original digitised images after application to the owner are free to carry out their own restoration process. The DIAMM project agreed that all restoration is a form of editing; the scholar executing the process makes subjective decisions about how best to highlight the faded ink. Each attempt at restoration brings out subtle differences and no two restorations are the same. Thus, rather than imposing their editorial decisions on others, the DIAMM team are happy to let other scholars perform their own restorations, while also offering their own work where that would be more useful.

The DIAMM has an obvious appeal for musicologists keen to analyse manuscripts previously believed to have been lost to time. Bringing together the collection also allows the corpus to be considered as a whole. Before the existence of the website, many scholars avoided incorporating manuscripts into their research that may have necessitated a travel and/or expensive purchases of photographs and microfilms. With a manuscript now online, it can be compared to images from, for example, Aberystwyth, Tours and Rome. Equally, scholars residing at a distance from European archives can now examine the manuscripts from their own home. But it is not only professional musical scholars that make up the 30 to 60 daily hits that the website receives. Historians, codicologists, undergraduates, school pupils and early music ensembles have all made use of the extended access provided by the digital collection. The technical skills learnt by Dr Craig-McFeely have also been in demand, and more archives are now using these techniques in conjunction with their own digital equipment. Most recently, the project was approached to provide high resolution 'cleaned' images to a scholar writing a software program that would recognise neumes (signs representing a group of notes in plainsong scores) in the same way as Optical Character Recognition software enables text to be scanned as text, rather than as an image.

Conclusion

The Digital Image Archive of Medieval Music team is currently proceeding with the second phase of the project, which involves the digitisation of fragments of English origin, mainly held in archives in the UK, but some on mainland Europe. From August 2001, phase 3 comes into operation, which will embrace all medieval music fragments, including those of non-English origin, and will also allow the digitisation of many complete manuscripts. The team also hopes to uncover some previously unknown fragments - five have been located already. Once the project is complete they will have digitised over 5000 musical fragments, and most of these should be deposited on their website (copyright permitting). The team is also negotiating for the right to disseminate a greater number of the manuscripts without password controls. But it is their ambition to have a website containing images of the scores before and after their visual restoration that is perhaps the most appealing aspect of the project. While many digitisation projects concentrate on preserving documents and providing access, the DIAMM project adds to this by providing 'new' documents that contemporary scholars have not previously investigated. Such digital innovation allows scholars to conduct a whole new range of musical research.