This article discusses the Digital Image Archive of Medieval Music (DIAMM), and the ground-breaking work they have done in digitizing manuscript sources for medieval music. In many cases this has involved digital reconstruction where the notation is illegible, erased, or hidden under later texts. Through their good example DIAMM have been instrumental in educating photographers in archives and researchers alike in the proper methodology of digital photography with regard to medieval manuscripts. Their standards are especially high in capturing the images and the metadata they store. This enables them to undertake the sometimes miraculous digital restoration. The article covers the methodology of the DIAMM project, the delivery of images, and some of the problems they have encountered.
In 1998 two medieval musicologists at Oxford and Royal Holloway, Margaret Bent and Andrew Wathey, started work on a facsimile volume in the long-running series,
There were a number of reasons for this decision: firstly the collection was of fragments, often in poor condition, ranging from pieces the size of a large postage stamp to several complete bifolia. These were in danger from natural decay, damage hastened by medieval 'vandalism', and poor husbandry. Some which should have been included had been stolen or mislaid. Study of the repertory as a whole was nearly impossible due to geographical spread, and hampered further by the often appalling condition of the sources. The damage evident was almost all due to the re-use of the parchment, but in some cases resulted from early 20th-century attempts at restoration. Some examples of the types of challenges to transcription are seen below: the first is the lining of a hat box, the second was used to mend the inside of an organ case, the strips had been used as quire-guards; the single surviving leaf of what was clearly an extremely opulent choirbook was eaten away by rats and mould, and the final group were scraped off a ceiling, where they had been used as wallpaper backing.
Bent and Wathey consulted with Marilyn Deegan (principal technical advisor to DIAMM) and found that new digital imaging and image-processing technologies might enable them to manipulate the images they acquired, if they were of high enough quality, to improve the readability of the texts, and thus produce a volume of photographs that was more useful than a conventional facsimile.
Secondly, they believed that the cost of creating colour digital images would be the same as producing digital black & white, and since much of this repertory utilises different-coloured inks to represent different rhythmic interpretations, colour was essential.
A third reason for creating digital images came from their experience with their own collections of slides, transparencies and microfilms, gathered painstakingly over nearly 40 years. These media were deteriorating; even relatively new slides had discoloured, and every time a microfilm or fiche was used it became slightly more scratched. Their everyday use of computers suggested that accessing images as digital objects would be far easier and more productive than conventional means of studying surrogates.
Given the fragile state of the repertory to be studied, and
the imminent loss of much of the information it contained,
they felt that a permanent archive of some sort
In 1998 Bent and Wathey saw no particular need for a website (research websites at this time were virtually unknown). Once the project began though, it was decided that a digital project should have an online presence, even if it was only a brief front page. It is a measure of the incredible speed with which the internet has become a serious academic resource, and the importance it holds now, that as recently as 1998 a website was considered frivolous, and 'archiving', as described at a British Academy conference in late 1998, was discussed by many academic projects in terms of how to store your word files, whereas now a website is considered an essential part of any project in reaching the wider research community.
Failure to participate in the digital world is now frowned
upon: those without access to the internet are actively
discriminated against ("10% discount online" is a familiar
advertising catch-phrase); academics who fail to participate
in the digital medium, or who refuse to use digital
resources are in danger of being labelled dinosaurs, and
missing out on vital information or discourse, because now
that is the
Technology gradually impacting on the humanities began as the driving force behind many academic digital initiatives such as DIAMM; now technology is having to move to meet the demands of an increasingly technically knowledgeable academic community. When DIAMM first started digital imaging, a large scanning back camera existed, but computing available to projects such as this was not sufficient to manipulate images of the size that it produced (280 MB). Therefore, instead of buying the top-of-the-range camera, the project started with one a step below, which produced 80 MB images, making it practical to consider software manipulation with the resulting images. Within a few years computing had caught up with the scanning backs, and the equipment was upgraded to the larger camera, with a corresponding increase in both the quality of the images, and the complexity of the digital restoration that could be undertaken.
Storage has always been a critical component of DIAMM, and the project has been extremely fortunate to have the support of, and access to, the hierarchical file-server managed by Oxford University Computing Services (OUCS). At present our content occupies about 10 Terabytes of space. Using uncompressed TIF format for the images increases storage needs, but is preferable to a lossless compressed format that may not be readable by contemporary software in a few years time, or may require a migration process that alters the data in the image. Since digital archiving is relatively new, we still do not know whether our file formats will have the longevity we hope for, but at present we follow best practices and keep a weather eye on technology and software developments.
Grants from the Arts and Humanities Research Board (AHRB, formerly the Humanities Research Board) in the UK have enabled DIAMM to exploit its now-extensive expertise in the field of high-resolution digital imaging and extend its remit to include a broader range of fragments than the original group, and also embrace the very considerable corpus of complete and comparatively undamaged manuscripts surviving throughout Europe, not just those on our doorstep in the UK. Since the imaging equipment is completely mobile, and our protocols are well established, we are able to produce images of completely consistent quality in archives as widely separated as London and Hikone in Japan, and we have worked in archives throughout Europe.
We have digitized manuscripts for a number of other
projects and individuals, musical and non-musical, medieval
and modern, ranging from 2nd-century Chinese scrolls to
Medieval mystery plays, Anglo-Saxon Charters to Jane
Austen's holograph (a representative list is available on
our website:
Image quality is still a major issue in digital imaging, since so many suppliers are still producing digital images of appalling quality, believing that this is all that digital cameras are capable of. Unfortunately these are not amateur organisations: they are often the photographic departments of major international research institutions.
Several institutions have published information about
imaging standards:
In spite of the availability of this sort of standard and the reasons behind its creation, there is an extraordinary level of ignorance regarding quality, and a surprising inability to evaluate digital images and see problems which should be obvious. An alarming number of institutions are digitizing at spectacular speed, but still have not attempted to calibrate any of their equipment, so none of their images have accurate colour.
One of the projects to which DIAMM acts as an imaging advisor put a series of orders for high-quality digital imaging in to a group of international libraries. Only the British Library managed to meet the imaging specifications, which were basic and simple, and designed to create a corpus of consistent images to facilitate online comparative use:
- Images should be taken at 400 dpi resolution at actual size;
- A colour scale and size scale must be included in each image;
- The image must be saved in uncompressed TIF format, with no JPG or other compression format used at any point during the capture process;
- The colour profile of the capture device should be embedded in the image;
- No unsharp mask, level adjust or any other process should be applied to the image either during capture or after;
- The picture must be in focus at pixel-for-pixel view.
There are other specifications which we would add in an ideal world, but we have found that just getting the supplier to meet this baseline standard for an acceptable digital image is extremely difficult, so it is a waste of time to ask for more.
Despite clear quality specifications, the images received in response to the orders were rarely acceptable in quality: one library captured everything at 200 dpi, then used image-processing software to 'increase the resolution' to 400 dpi: the result was a blurred image. In order to increase resolution in this way, the software has to interpolate new pixels between the ones that already exist. This is done by inserting a pixel with a colour value halfway between the colours of the pixels on either side of the new one. If the original image has a white pixel adjacent to a black one, the interpolated pixel will be 50% grey, softening the previously sharp demarcation.
Another library saved the images in compressed JPG format,
so that all the delicate gradations of colour were lost: JPG
is a 'lossy' compression format, so it destroys data by
storing colours that are
In the following segment of a b/w image, JPG compression has been used for successive saves. The first image is as the shot came off the camera (which unfortunately stores to JPG by default, but at high-quality compression). The point to note is the clear gradation from dark to light grey across the shot (this is part of a beach – the rest of the image has been cropped off to save space). The second image has been subjected to low-quality JPG compression. It should be visible that the gradation in grey has now changed from smooth, into blocks with sudden shifts from one tone to the next. The third image has had a further level of compression used, and this is now almost unintelligible: the gradation is almost entirely lost, 'blocking' has appeared, and there is a 'watery' look to the picture caused by the loss of detail.
(If you find that you do not see a smooth grey on the first picture, check in your monitor settings that you have your display set to 'millions of colours' or high resolution. If it is not, changing it will dramatically enhance the appearance of everything on your computer screen. Most of the samples in this article will not be informative unless viewed at high resolution.)
Archive quality digital images should not use JPG at any
point in the capture or delivery process since it is a
destructive process, and colours discarded by the
compression algorithm
The colour of many images is often incorrect in some way – it is not unknown for the operator to correct the colour appearance of the image while viewing it on an uncalibrated screen. All uncalibrated monitors have a colour cast, so the operator was compensating for the colour deficiency of their own hardware, and falsifying the colour information in the image, so that when viewed or printed on properly calibrated equipment, the image had a colour cast. Other colour problems have been caused by embedding the wrong colour profile in the image, so the software displays the colours as they would appear if scanned on one piece of equipment, but the picture was scanned on something completely different. Fortunately, where the archive complied with our request to include a colour scale, we are able to see where there is a colour problem, and it is often simply the misapplication of a profile.
The following examples are copies of the same image. The first has the correct profile embedded: it should have a grey border, but your screen may be mis-calibrated (most are: they come out of the box like that). Unfortunately the human eye compensates for colour casts: if you put on a pair of green sunglasses in bright weather, the sky still looks blue, even though it is not, because your brain is compensating for the superimposed colour. This is called chromatic adjustment. If you are used to viewing the world via a poorly calibrated screen, the grey border of this image may look grey, but if you hold a grey card up to the screen, you should see the difference (assuming there is one).
The second picture is the same image with the wrong profile embedded. If nothing else you should be able to see that it has changed colour. On a correctly calibrated screen the colours in this second image would appear heavily saturated and the grey border will appear pinky-grey.
(Simple instructions for checking and calibrating your
monitor may be found on the DIAMM website:
One institution found that nearly half of its camera operators were colourblind to some degree, but they had never been tested, and were not aware that they had colour-deficient vision. This should not affect the ability to take a good picture, but would make evaluation of colour liable to error.
(To find out if you are colourblind, take these quick
tests:
More disturbing is the propensity of the operator in our target institutions to fail to get the camera properly in focus before shooting. The result is a soft-focus image, which the operator believes 'can be corrected in image-processing software' afterwards using the Unsharp Mask. If the image had been in focus to begin with it would not require any post-processing time, and sharpening simply serves to put a bright corona around all the marks on the page, including dirt or mould, and increases graininess, all of which contribute to the image being considerably less useful than it should be. An increase in graining, and resulting 'flatness' of the picture is one of the imaging flaws that many suppliers seem to find difficult to see.
When DIAMM first started taking digital images and
approached libraries to digitize items in their collections,
several librarians said that they had seen digital images,
and didn't like them, so were reluctant to let us digitize
their holdings. Unfortunately, they had only seen bad
images. There are still widespread misconceptions about the
sort of quality you can get from a digital image, and these
are often due to a very poor understanding of how digital
images work. For example, many users of images do not
understand that reducing the size of an image to e-mail it,
means that it is no longer large enough to print at high
quality at that size. The relationship of the screen image
to the printer is ignored. The dots that any printer
produces on a piece of paper to make up an image are far
smaller than the grid of minute squares that make up a
computer monitor – at least three or four times smaller. We
view screen images at 72 dpi (dots per inch) or now more
usually at 96 dpi (as screen quality has improved), but to
print you need the image to be at around 300 dpi, or the
dots will not be close enough together to create sharp,
clear pictures. A high-quality digital image, whether
displayed on screen or printed can, and should, be
The next example (I will not embarrass the supplier by identifying it) shows part of an image that has been sharpened excessively to compensate for focus shortcomings. It is grainy and in places (elsewhere on the page) unreadable. If this is the quality the supplier expects from their images, then it is hardly surprising that they consider digital a poor substitute for analog.
The next image is acceptable in quality, apart from some colour banding along the edges of the black lines indicating a fault with the scanner that the library probably hasn't noticed (or more likely in this case cannot afford to fix). It had no colour or size scales, so it is not possible to evaluate the colour, but the quality of the scan in focus and clarity is excellent.
Another flaw in imaging that is generally corrected post-capture, is poor lighting or incorrect exposure. An over-dark picture can be lightened using a tool called level adjust. However, doing this, which causes all the colours present in the image to be 'spread out' creates gaps in the colour spectrum represented, as well as moving the colour values to new values, so the result again is a falsification of the information in the image. The first histogram shown here is of an uncorrected image, the second shows the gaps in colour values caused by stretching the values out.
The sort of imaging undertaken by DIAMM is not in any way comparable to consumer-level digital cameras, although the gap has started to close in the last year. A good high street SLR camera is now capable of taking a picture at 8–10 megapixels (The new Canon 400D digital SLR takes 10 megapixel images). This will produce an image that can be printed at A3 size at reasonable resolution, so the image still looks clear and sharp in the details, even at that size.
DIAMM uses a PhaseOne PowerPhase FX scanning back, mounted on a custom built focus box supplied by ICAM Archive systems, specialists in archive imaging equipment. Images have a maximum capture area of 144 megapixels and file sizes are in the region of 280 MB (NB, not KB), whereas an 8 megapixel camera would produce 22 MB images if saved to TIF. In order to archieve this sort of resolution, current technology uses a camera with a digital scanning back on it, since the cost of a sensor that size would be prohibitive – such sensors do exist, but they are used in spy satellites rather than general-user technology.
A scanning back is not unlike a miniature flatbed scanner: it has two rows of sensors which correspond to a quarter of a pixel in size for each sensor element. Thus the FX uses one row of 24,000 alternating red and green sensors, and another row of 24,000 alternating green and blue sensors. Each group of four pixels (red-green/green-blue) captures an image of one pixel in size in the final image.
(There are twice as many green sensors as the other colours, because green is the most difficult colour to see and to capture.) The scanning back takes a strip image 12,000 pixels wide, then moves forward and takes another strip, repeating the process a maximum of 12,000 times. The result is glued together into a complete image by the capture software.
The main disadvantage of this technology is that each picture takes a long time: one scan can take in the region of 5 minutes, whereas with a single-shot camera back, capture is instantaneous. High-resolution imaging is therefore very costly in comparison with lower-resolution or analog alternatives, although it still has the advantage over analog that there is no requirement for film processing or printing.
Until recently there was a huge gap between the quality available to the consumer market, and that used by archives and by DIAMM. Only recently has the demand for bigger and better imaging in the hand-held single-shot market given birth to a new generation of professional digital capture media exemplified by the 39 Megapixel single shot camera backs produced by Hasselblad and PhaseOne. However, these are still highly specialist professional equipment: one of these digital sensors, with a camera and lense(s) of appropriate quality on the front costs in the region of £25,000–£30,000.
For those who are interested, PhaseOne have announced that they will not be developing any further scanning backs, and future development will now be towards single-shot capture equipment.
To clarify the type of images the various cameras produce, a snapshot of a page of an Arthurian manuscript from the John Rylands Library in Manchester is given below, photographed for Dr Alison Stones' Lancelot-Graal project. Dr Stones, an art historian, was interested in the historiated initials on each page. The MS page is quite large: if we had photographed using a Canon EOS 350D (8 megapixels), this is the maximum size the miniature would appear on screen, without enlarging beyond a one-to-one pixel-to-pixel relationship:
Photographed with a PhaseOne P45 39 megapixel single-shot camera the on-screen image would have appeared like this:
For many users that would be more than sufficient. However, with the project's main camera, the PhaseOne FX, one of the largest of the scanning-back generation of digital sensors, the detail possible, particularly in on-screen view, is considerably greater:
What is the purpose behind imaging at this extraordinary
quality? The main reason had its origins in the group of
sources for which DIAMM was original conceived. There is a
considerable corpus of fragments of medieval polyphony
distributed around the world: access to them is often
difficult, and if the fragment is very small, the cost of
seeing it in person may be too high. Gathering these sources
together as b/w glossies or microfilms is a costly and
difficult process, and the outcome does not really give the
scholar materials that are good enough surrogates for
complex research, particularly if the original is damaged,
which it nearly always is. Part of the remit of DIAMM is the
reunification of a corpus which has become very widely
scattered over the centuries, and the provision of images of
a quality that will facilitate study of the document that is
significantly
We take images directly from the original source, and not from good surrogates such as Ektachromes or colour glossies, since our resolution usually far exceeds that offered by any surrogate. The following samples demonstrate the difference between a scan of the original document and a scan at similar resolution of a good colour photograph.
The first difference is the colour: the photo scanned for the left image was a few years old, but did not include a colour scale, so we did not know that it had changed colour with age. (It is possible that it had always had a pink cast and hadn't changed with age, but without a colour scale included in the picture we could not tell). We also didn't know if the document had changed colour itself by the time we reached it, as a colour scale would have supplied that information (which is one reason that we insist on including scales in all pictures). You may be able to see already the difference in sharpness of the two images. The one on the left used a UMAX high-resolution flatbed scanner to scan the photograph, the one on the right was taken with the DIAMM PhaseOne PowerPhase, imaging the source directly under daylight balanced lighting conditions.
When we enlarged the two scans (which were made at the same resolution) the photograph became too fuzzy to be useful very early on. Our digital scan however, because it was properly focused, stayed crisp and readable right up to one-to-one resolution on screen.
The fragments in our original remit only survive today because they were re-used when the music went out of fashion. Parchment and vellum had intrinsic value that meant it was recycled in many different ways: at best it might be used as a wrapper for other documents (in which case it might be in reasonable condition); worse scenarios (for the music) are to be found when the surface was scraped, refinished and written over, so that the music is palimpsest. Even more damage was caused if it was used as paste-downs or strengthening for bindings. Some of the items photographed are barely recognisable as parchment. Given the parlous state of our core starting corpus, the extremely high resolution was essential in order to examine the manuscripts in fine detail – far finer detail than can be seen with the eye, even with a magnifying glass. With this sort of quality and colour separation, it was possible for us to develop digital restoration techniques which have returned to legibility documents for which the text was believed permanently lost. It is this activity for which DIAMM is best known in the musicological and wider manuscript study community.
Most image restoration activity is centred on the restoration of damaged photographs, where the surrogate, rather than the original, is the object of interest (e.g. Disney's famous restoratin of their Snow White). Most of these deal with historic photographs, but some have a more serious application, such as improving medical imaging for better diagnostic capability. Restoration of historic photographs and glass plates has yielded a wealth of historical data, but more recent history is also relevant, since there are companies with specialise in 'improving' early digital images There are a few projects involved in using digital imaging as DIAMM does, to improve the visibility of real objects. Scientists working at NASA contacted DIAMM early on, to see if our work was complementary, but their main interest was in developing repeatable algorithms. Exchanging images and trying our various techniques confirmed that repeatable techniques would not work on DIAMM sources.
There is a wealth of information on the web regarding these types of restoration to be found by searching on the web. A few examples are given here.
Basics of our restoration techniques are described on the
project website,
Digital restoration was certainly a goal of the project,
but creating new software to do that was very definitely not
something we wanted to become involved in. The Centre for
the Study of Ancient Documents in Oxford (
Two commercial packages, Adobe Photoshop and Paintshop Pro were considered, but Photoshop won out mainly because (at that time) it offered something which no other software did: the ability to create and save layers of work, much as transparent overlays might be used with an overhead projector. The file sizes grew with every overlay, but their use did mean that processes applied to the underlying document could be turned on and off, shuffled, or adjusted in different ways.
Photoshop offers a vast array of tools, most of which are ignored in our restoration processes. However the power of the software underlying those tools is essential for the aspects of it that we do use. It is developed as an artistic tool, and one which enables professional photographers to do in the digital medium what they used to do in the darkroom. It was certainly not conceived for digital restoration, though it does the job extremely well.
Key to restoration is the ability to select very specific colours: JPG compression attenuates the colour spectrum in an image, thus limiting the quality of restoration that is possible. If an element in an image is enlarged sufficiently, it is possible to see that what might appear at first to be a black or brown mark, is in fact composed of a very large number of colours, which can be separated out when enlarged sufficiently. Photoshop is able to differentiate electronically between colours which the naked eye cannot perceive a difference in, and therefore this allows the user to select colours which are all but invisible, and darken them to increase readability. It also allows the separation of colours which are nearly the same so that, for instance, palimpsest text can be separated from text written over it in ink of very similar colour.
Most of our restoration work relies on the ability to separate and define colours very accurately and in minute detail. Once you have done that, there are a variety of simple processes or tools which can be used to darken or lighten text or dirt. Sometimes it is not even necessary to select colours before using the lightening/darkening tools. Most superficial dirt can be faded back, and underlying ink brought to the fore by simply using the level-adjust tool.
Techniques of image restoration is a vast subject, and it
is not possible to describe it in any detail here. As well
as the brief description on our website, DIAMM has published
an Image Restoration Workbook, written to accompany a
workshop where restoration techniques were taught. It may be
downloaded (free) from
The first example comes from the Shakespeare Birthplace Trust. This bifolio is used as a wrapper around more delicate paper legal documents. On the inside face the music is clear and clean, but the catalogue description of the outer side describes it as having no music visible. Restoration has revealed a transcribable texted secular song: the result is not intended to restore the manuscript to its pristine state – that would probably not be possible without considerable editorial intervention (or 'faking up'), but it has rendered the content readable to a relatively inexpert user level.
A rebinding programme in the 1960s and 70s in Cambrai resulted in the binder discarding the original endpapers of a number of manuscripts, papers which preserved a lost musical repertory from a dismembered manuscript. A number of leaves did survive thanks to changes in policy in the restoration bindery, but many are now only known as offsets on the original oak boards. The digital image is flipped to create a mirror version of the offset, then the dark or colour writing is separated from the colour of the wood and leather boards.
The British Library and Bodleian Library manuscript and early printed book collections are particularly rich in endpapers, the one shown below, from the British Library was trimmed to size to fill out the binding shape left unoccupied by the leather turnovers. Layers of glue and other dirt concealed not only music, but a significant section of text, which has now been transcribed.
Corpus Christi College, Oxford has a well-known collection of medieval manuscripts, among them MS 144, containing the poems of Geoffrey of Vinsauf. Vinsauf's text is easy enough to read, but must be digitally removed in order to reveal the musical palimpsest, which was discovered since the galls in the original ink had left traces which were becoming more visible with time.
There are ethical consideration in restoration: the tools available in Photoshop allow a type of restoration that relies heavily on editorial judgement. In the next example, the damage caused by writing on the reverse of the leaf burning through the paper has been 'cloned' out, by replacing the damaged areas with segments of undamaged parts of the page (the third image shows the cloned 'patches' which have replaced ares of the original, showing the extent to which areas the document is no longer a true representations of the source). Where this is just a case of eliminating material which is obviously show-through, or burn-through, there is less likelihood of introducing errors. However there are places where the editor 'repairs' damaged musical notes, replacing them with what s/he believes should be there, and that may not be correct. In which case, the result is misleading. This particular document could be repaired much further, but this is about as far as the editor can go without making decisions which cannot be based on what can be seen
This is the type of virtual restoration undertaken by
Fotoscientifica, a commercial organisation in Parma, Italy
(
The main concern of our depositors was not that their images might be stolen, but that their sources might be misrepresented in some way. This led us to change our mode of restoration: previously we had attempted to restore using 'naturalistic' colours similar to the original inks, mainly by darkening or lightening particular colour selections, but this could be mistaken for the actual appearance of the source in some cases. Restorations on a Florentine complete palimpsest manuscript was found to be far more successful when the material that we wanted to restore to readability was coloured an unlikely colour such as green or purple. Although the same colours were selected in each restoration, the coloured results were far more readable than those using natural browns, and this technique had the advantage that there was no longer any possibility that someone could mistake a restored version for the original.
DIAMM has become a significant collaborative effort between Medieval scholars and those with technical expertise, resulting in the creation not only of an image archive of exceptional quality images of European medieval music manuscripts, but a delivery system that allows the research community, and other users, to access these images with ease.
Early on in the project scholars discovered the offline archive (established purely for preservation purposes) and started to ask for access to the images, preferably on the internet, which was fast becoming a natural means for communicating data without distance limitation. In 1998 internet resources in the humanities research community were very limited. Many of the libraries whose documents we had digitized did not have access to the internet at all, and were naturally very suspicious of this medium. The rights in the images we had created remained with the owners of the documents, a policy which was in some cases solely responsible for the owner agreeing to allow us to digitize their materials. Despite initial misgivings every one of the UK libraries, and many of the European ones that we asked to allow their images to appear in our online resource agreed.
Suddenly a corpus that had only ever been studied in isolated pockets, and usually only by senior scholars who had the finances and commitment to the corpus to gather surrogates for themselves, could be studied by anyone – academic or not – from their desktop.
The first website was designed only for a small number of sources, and using the best technology available at the time (PDF – portable document format) that would allow zoom and rotate functionality. In order to get moderate resolution, the user had to wait for the whole PDF to download before they could view the image, several minutes in some cases, particularly with dial-up access which was then the standard for non-university spine sites. The Andrew W. Mellon Foundation funded a major scoping study to develop a system that would allow us to deliver high-resolution images at speed, as well as accompanying them with metadata that had been absent from the original website, since it was originally only intended for specialist users.
Several successive grants from the Foundation have facilitated the development of a feature-rich web delivery system for our image collection, and the expansion of the metadata resource beyond only those manuscripts in the image archive. Web-delivery is managed by the Centre for Computing in the Humanities at King's College London, where browser technology is constantly pushed to its limits to enrich the online research environment in a number of music projects. As an academic department they are committed to open-source development, but if a piece of software does the job better than anything else it will not be ignored simply because it is not open source.
Most of the website is accessible to non-registered users: project information, notes on image restoration, library address lists, the source lists and metadata, as well as access to page-images of the printed catalogues and electronic versions of the catalogue texts. Interested readers may consult these parts of the website at any time: www.diamm.ac.uk.
The most important aspect of the new development was the implementation of the Zoomify viewer (www.zoomify.com) to display the images. We are now able to present registered users with the full size (up to 320 MB) images, which download instantly, and can be zoomed and panned in a full-screeb window with little appreciable time delay. (The full size images may only be accessed by registered users, due to rights-protection requirements of depositors.) A number of image-based resources now use this software, including most major art collections and auction houses worldwide.
The size of image viewed, in the case of DIAMM, is limited only by the user's screen size. Several windows can be opened simultaneously, allowing side-by-side comparison and, thanks to consistent imaging standards, true comparison is possible in this medium. A recent adjunct to the image-viewer is the list of 'secondary' or 'alternate' images which now appears in the toolbar. These are UV, watermark (using a light sheet) or restored versions of the same page, and clicking on the link brings this image up in the same viewer so that they can be compared side-by-side at similar or different magnifications. The next phase of development will see much wider exploitation of this tool, which has considerable features which we are not yet using. One department which has experimented with the possibilities in Zoomify is the University Of Melbourne's Educational Technology Services. This is by no means an exhaustive selection, but it gives some idea of the flexibility of this tool.
Zoomify is not Open Source. Neither, in all probability, is the browser being used to read this. Nor is the word processor or PowerPoint software that are relied upon so heavily for everyday work. Some open-source Java-based viewers have been developed which claim the same functionality as Zoomify, but are much slower, since Zoomify uses the Flash plugin. Our decision to use this non-open-source software is based on functionality and the increased availability of our resource to the end-user that it provides. Since Flash is now installed as standard in most browsers, the user does not need to download anything to access our images. Since this part of the website is not viewable to non-registered users, I have provided some screenshots to demonstrate some of the facilities available.
Workshops with musicologists led to the design of several tools to accompany the images: the first (designed principally by John Bradley at CCH) allows the user to create personal annotations which attach to an individual image, persist between login sessions, and are not visible to other users. The tool includes a small set of formatting commands (activated by clicking an icon), and the facility to paste in source or image reference numbers from other windows (the numbers are given beneath each image) which become live links to open other source descriptions or images. Extending this tool, we provided a nearly identical tool, but one which is visible to any user of the site, thus creating the facility for an open discussion forum, though not on a Wiki model (suggestions that we implement a Wiki forum are being considered in the next phase of work). Finally we added a text transcription tool based on the same engineering as the commenting tools. We are gradually adding full text transcriptions for all the sources in the collection that will eventually be fully searchable both in original and standardized spelling; this tool allows users to contribute to the work of transcription, and their contributions will be moderated before incorporation into the search system.
The location of any comment made by a user is saved in a 'MyDIAMM' area, which only the logged-in user can see. This creates pick-lists (like a shopping basket) of images or manuscripts which circumvents the original search or browse process necessary to locate an image.
Where there are secondary images that supplement an original, such as ultra-violet or restored versions of a leaf, a further tool appears on the palette, offering a list of secondary images which, when clicked, split the main viewer to leave the original image on the left and present the secondary one on the right. The two images can be panned and zoomed independently.
The database behind the online resource has grown immensely in the last two years. Originally it was an administrative tool that allowed us to keep track of our imaging work, and store sufficient metadata that we would know when on site whether we had been given the correct manuscript. Now, it is the repository for a massive metadata resource covering all medieval music sources, not just those we have photographed.
The database has been populated directly from the contents of two multi-volume printed catalogues which are normally only available in libraries. Although the content is fully integrated into the database, we have provided users with the facility to access page images of the original catalogues, and browse any volume page by page. We have found that the website is visited more frequently by scholars accessing the catalogues than by those searching for images, highlighting a user requirement that had not been anticipated, nor specifically raised in our workshops or user-group surveys.
The search engine is currently constructed on standard principles, but relies on a source-based access method. For users who are not familiar with searching sources by the library where the manuscript is housed, the search engine is being expanded and will cover a much broader set of criteria that should enable users to create the sort of search results that are not possible outside the digital medium.
At present this is an image-based resource, but the next phase of development (2007-9) will transform this into 'an information resource with images'. Users will be able to create inventories for manuscripts (surprisingly something that is not available for all catalogued manuscripts), source lists for composers that cross the boundaries of single manuscripts, and other personalised research materials which will be retained in their personal workspace between login sessions, and may be made available to other users if required.
Online Registration will be in place in early 2007, but presently creation of a user-account still requires the completion of a hard-copy access agreement which has to be posted to the project manager. At the moment the resource is free of charge, but one of the conditions of our funding is that by the end of 2009 we are self-sustaining, and inevitably that is going to involve the implementation of some sort of charging model – probably for the research metadata and tools: we intend to keep access to the images free.
DIAMM has become a significant information repository, and over the years has developed close ties with other musicological projects overlapping or dovetailing in content and repertory. Plans for the next phase of work include the collaboration of a number of libraries which house books originally created in the Alamire workshop in the Netherlands. The Alamire books are a famous example of a nearly complete corpus of richly illuminated manuscripts, which now belong to libraries in Jena (the largest group), Vienna, s'Hertogenbosch and elsewhere. DIAMM will be bringing these collection-holders together to establish a project that will seek funding to digitize these sources and virtually re-unify them through the DIAMM website, where they may all be accessed, even though the original images will reside on the server of the participating institution.
An evolution in metadata management is about to take place
with DIAMM and four other major metadata creators in the
musicology field: Oliver Huck's Die Musik des
Trecento database and variorum representation project
based in Jena and Hamburg; Theodor Dumitrescu's Corpus Mensurabilis Musice
Electronicum in Utrecht, Thomas Schmidt-Beste's
motet database in Heidelberg and the University of Bangor
(early samples of this can be seen at
These projects plan to create a single collaborative
database, dealing not only with text metadata, but also with
newly created
The project team is particularly anxious that the resource they have created should be exploited more widely. We have therefore invited a number of projects for which we have done imaging work to display their images through the DIAMM website. DIAMM has 'unlimited' server space, a facility rarely available to collection holders or small projects, and the website provides a rich research environment in which to present images. We are therefore actively soliciting deposits from other Medieval projects who need a delivery system, but whose funding does not permit them to develop a system as complex as that offered by DIAMM. It will be possible to define independent projects within DIAMM so that it retains its identity, but the tools and features will be consistent across the resource.
At present there would be no charge for inclusion, but the depositor must provide a certain level of metadata to accompany their images, and must negotiate the rights for online delivery with the document owner(s). (If you need assistance in negotiating rights please contact DIAMM for advice.)
DIAMM is held back by the limitations of the web but propelled forward by the needs of our community. In providing this resource, research methods have changed, and continue to evolve pro-actively and in response to the potential of the online medium. The rapid advance and emergence of new web and digital technology is a constant challenge to long-term planning and sustainability, and we rely heavily on our technical partners at CCH to keep the resource we have created from stagnating. Our intention was to create images that would stand the test of time, and so far they are doing that, although we are still faced with the unknown of digital longevity, storage media and file formats. The project has expanded in every direction since its inception, and continues to do so both in response to technology and to the needs of the wider research community that it serves.
Digital imaging has been around for quite a while now, leading to a false impression of knowledge or skill, but is really still in its infancy: too many archives are setting up imaging with little or no expertise and without the appropriate backing in such basics as colour calibration: their staff cannot tell the difference between a good image and a bad one; in some cases they rely on outside suppliers who QA their own work and tell the library that they are getting something good (which they probably believe), when in fact they are getting something appalling. Recently someone involved with a research project remarked that they were thinking of buying a digital camera (a PhaseOne P45, so not an idle outlay) and starting to take pictures of manuscripts themselves, despite the fact that they had no idea what was required to take a good image. There is still the perception that archive imaging is merely a question of pressing a button (like holiday snapshots), and the ghastly consequences of this complacency are to be seen all around us.
Since its inception DIAMM has had to swim against a tide of misinformation presented in the guise of 'expertise', or quality expectations based on poor exemplars. Particularly frustrating for DIAMM is that an archive will often refuse to allow better quality imaging to be undertaken if a manuscript has already been photographed, thus leaving an artefact to deteriorate with no accurate, preservation-quality record having been made of it. In some cases, the results produced by a supplier have put off the archive from ever having any further imaging done, and they assume that DIAMM produces the same poor images that their first encounter produced.
Although there is nothing that can be done to address the problem of visual acuity in evaluating digital images, a number of projects and digital image producers recognise that the lack of a universally-accepted standard is a barrier to progress in improving imaging quality, particularly with documents for which there is perhaps only one chance to get a digital image. In collaboration we intend to produce a standard that can be disseminated with the backing of major institutions to establish benchmark procedures that will ensure a certain level of quality for all archive imaging. The basic 'rules' are listed above ("Image Quality"), and we hope to publish and disseminate our paper on standards during 2007.