Category: Digital resources

Up close and 3D: photogrammetric models at the Bodleian

John Barrett (Bodleian Imaging Studio and ARCHiOx Project) introduces the Bodleian’s growing collection of photogrammetric models

Explore on: https://sketchfab.com/ARCHiOx/models 

Holy Bible; Psalms, 1660-1661, London. Embroidered binding. English, mid-17th century. Blue silk binding, embroidered with silver wire; portrait of Charles II on front board, Catherine of Braganza on back.  Broxb. 45.2. https://sketchfab.com/3d-models/broxb-452-8fbcd617f9bd4e3799bbad8ce7d54add

Since 2022, the aim of ARCHiOx (Analysis and Recording of Cultural Heritage in Oxford) has been to record the three-dimensional surface of items in Bodleian collections.  The recordings have demonstrated that the micro-topography of books, manuscripts, printing plates and artworks can tell us a huge amount about who made them, how they were made and about people’s interactions with them since they were made.  An imaging system called the Selene, developed by project partner the Factum Foundation, has been used to acquire this 3D data.  Using a principle called photometric stereo, the Selene is able to record relief on the surface of an original with a maximum variation in height of around 15-20 millimetres.  This makes the Selene well suited for the recording of library and gallery material, where high resolution data is of great benefit and the originals are mostly flat.

Recording a book binding as an object so that every face is captured requires a different approach.  While the Selene acquires 3D data through capturing images illuminated from different angles and with the camera in a static position, photogrammetry is a technique which works in the opposite way.  A photogrammetric recording is made by photographing an object from multiple angles under even illumination and combining the images into a three-dimensional model.  In the case of the books and manuscripts which have so far been captured for the project, each binding has been photographed between 150 to 300 times and the images aligned in software.  The resulting recording can be navigated and relit by the end user onscreen, as if turning the original in their hands.

An early 17th century embroidered dos-à-dos binding combining The Whole Booke of Psalms and The New Testament.  N.T. Eng. 1630 g. https://sketchfab.com/3d-models/nt-eng-1630-g1-8aa0484213814a238fac3480e0e87f6f

A wireframe view of the same volume revealing the polygonal structure which forms the surface of the model.

Rather than moving the camera to capture the book from each position, by mounting the camera on a support and slowly rotating the book on a turntable, a series of captures can be photographed relatively efficiently.  Around 30 exposures are made per rotation at intervals of approximately 12-degrees.  The camera is then moved up or down and another round of images made until a set of captures are acquired from at least five different heights.  An additional two or three rounds of images are made after turning the book over to reveal the edge on which it was previously standing.  Depending on the proportions and condition of the volume, the book may be recorded horizontally, vertically, or both.  The surface of the resulting model is formed of millions of polygons on to which a texture is projected providing the colour and tone of the original.

In order to maximise accuracy and efficiency, a programmable turntable named the TablePi2 has been used extensively for the photogrammetry of the Bodleian’s books, manuscripts and artworks.  This highly efficient programmable turntable synchronises rotation and capture.   Using this solution, the bindings of small books can be captured in their entirety in just a few minutes.  The processing of the captured images into a 3D model is considerably slower and more complicated.

Thirsis Minnewit – De vrolykee zang-godin, 1690, Amsterdam. Dutch, late 17th century. Four volumes have been bound together back-to-back, in a style known as double ‘dos-à-dos’, in this copy at Bodleian shelfmark Broxb. 1.27. https://sketchfab.com/3d-models/broxb-127-240b716a1e2a4c28b3746edea02c0956

Preliminary recording of the same volume showing how a small change in the position of the upper board during capture has resulted in an unsuccessful 3D model with missing areas.

 While high-resolution surface data from an A5 size original can be recorded using the Selene in less than a minute, the capture, processing and editing involved in creating a photogrammetric model can take several hours.  Successful photogrammetry relies on the original not changing in shape during capture.  A change in volume of just a couple of millimetres can result in problems similar to the example shown above, where the upper board has moved between exposures.  Although a closed book may appear not to change considerably when it is turned either horizontally or vertically, the extent to which the opening of the fore-edge changes can often result in the images not being properly aligned.  For this reason, and in cases where the original has been recorded to reveal an opening, the lower edge may have been omitted from recording.

Oeuvres, Jacques le Royer, 1678, Avranches (shelfmark: Broxb. 46.10). This very rare volume of scientific tracts has a functional binding: one cover houses a stylus or gnomon that acts as a sundial.  https://sketchfab.com/3d-models/broxb-4610-dbd79fe9abe44eab9f688f3766903438

Where beneficial, volumes have been recorded so that both the outside and inside of the boards are shown in the model.  A note written by bookseller E. P. Goldschmidt, accompanying the volume shown above, states that this is an example of ‘A rare and extraordinary book printed at a small town in Normandy, preserved in its original binding which in fact comprises astronomical instruments without which no copy can be said to be complete.’ Before being presented to the Bodleian in 1978, this booked belonged to the Broxbourne library, compiled by book collector Albert Ehrman (1890-1969).  Ehrman had a particular interest in bookbinding and named his library after his hometown in Hertfordshire.  Many of the 3D models created and published by ARCHiOx have been recorded from volumes in the Bodleian’s Broxbourne collection, including the 17th century embroidered book of Psalms shown at the beginning of this post.  This volume features portraits of Charles II and Catherine of Braganza.  One of the first photogrammetric recordings to be made at the Bodleian, this model was made under the guidance of Ana Carrasco Huertas, PhD candidate, conservator and archaeologist from the University of Granada.

Dozens of 3D models have since been created and published.  Finding solutions for making successful models in an efficient way has involved a great deal of experimentation.  A recurring complication which myself and colleague Dylan Schirmacher encounter is in the processing of models captured from bindings bearing matching designs on both boards.  Often these have been so perfectly made that even the sophisticated photogrammetry software is unable to detect that the upper and lower boards constitute two separate faces rather than one.  The resultant model compiles all of the images to form a single face with four surrounding sides.  This is testament to the bookbinders’ skill in duplicating their design.

Auct. T. inf. 1. 10, New Testament, lacking Revelation (‘Codex Ebnerianus’). 12th century, beginning, Byzantine, Constantinople https://sketchfab.com/3d-models/ms-auct-t-inf-1-10-ce1a9803e7d2470484e4451fa2e13252

Reproducing the reflective nature of metallic surfaces is particularly challenging in both conventional photography and in photogrammetry.  The incredible silver and ivory binding from the Byzantine, Codex Ebnerianus was photographed from over 300 different angles to produce the model shown above.  By selectively masking only the silver parts of the model and applying a metalness mask, the author’s colleague, Dylan Schirmacher has produced a very convincing 3D recreation of the original.  Not only is it hoped that this recording will be beneficial to researchers, it will also safeguard this unique and fragile manuscript binding. 

[pr.] Shelley g.1, Sophocles, 1809, Oxford. https://sketchfab.com/3d-models/pr-shelley-g1-61a9301b60904911967f0ff6fe8df219

The majority of the volumes selected for photogrammetry have been chosen primarily due to their aesthetic quality.  The small volume shown above is an exception, and was recorded because of the intriguing story surrounding its provenance.  It was presented to the Library in in 1893 and catalogued by Librarian, E.W.B. Nicholson, as being ‘Found in Shelley’s hand at his death’.  The volume was displayed with the caption ‘Shelley’s Sophocles which he had with him when drowned.  Presented to the Bodleian Library, Oxford, by Jane, Lady Shelley [The end edge is supposed to show the mark of Shelley’s thumb]’.  Whether or not the story is true, there is good evidence to support the claim that volume belonged to Percy Shelley. The Bodleian holds a matching, pocket-sized edition of Aeschylus containing annotations in Shelley’s hand.  Had the Sophocles volume not been discovered with Shelley’s body, perhaps it could have been among the volumes salvaged from his boat, the Ariel.  What is undeniable is the presence of a thumb-sized indentation on the upper edge of this small, badly water-damaged book.

The reader is invited to explore the Bodleian’s growing collection of photogrammetric models using the following link.

https://sketchfab.com/ARCHiOx/models

This exciting project has been made possible through the generous funding support of the Helen Hamlyn Trust.

John Barrett is Studio Manager and Senior Photographer for Bodleian Imaging Services and ARCHiOx Technical Lead at the Bodleian Libraries

ARCHiOx, part 4: ‘Let him make a statue of a horse with its rider’

Camera photographing an ancient letter-seal
The Lucida uses a projected laser line and two tiny cameras to record the form of each surface of the seal. Bodleian Library, Sigill. Aram. V.

An essay by John Barrett, Senior Photographer, Bodleian Libraries, about discoveries from the ARCHiOx imaging project, which has been funded by the generous support of the Helen Hamlyn Trust. See also:  ARCHiOx: research and development in imaging – The Conveyor

By far the earliest collection of originals to be recorded for the ARCHiOx project originate from the Achaemenid Empire, and date to between 500 and 400BC.  The following image shows a clay seal, or letter-bulla, bearing the impression of the seal of Aršāma, a Persian prince and regional governor.  It is one of eight seals, which would have accompanied letters sent to the steward of Aršāma’s estates in Egypt. The impression made on this example, and six other bullae from the collection were made using the same cylindrical seal.  Lost to time, this incredibly intricately carved tool would have been rolled over the surface of each of these tiny clay seals, which measure little more than four centimetres.  The clay which forms these seals is unfired and consequently these small originals are incredibly fragile.  In some cases, the seals are held together by the string which would have attached them to the letters they accompanied.  Recording such vulnerable originals is of great importance to ensure their preservation.

A one-hundred-megapixel medium format digital camera has been used to photograph the four source images. In place of the custom flash modules, each seal has been illuminated using a studio flash unit.  The flash unit is moved to an equidistant position to the original at 90 degrees from the previous location, and the process repeated.

Recording the seals in this way has made it possible to capture them at over six and a half million pixels per square inch, but at this resolution the depth of field is extremely shallow.  Focus stacking is a technique whereby multiple images are photographed from a static position with an incremental adjustment made to the focus between exposures.  The resulting stacks of images are then combined in software. In this way the depth-of-field is extended and the recording appears absolutely sharp from top to bottom.  Perfect alignment of the four focus-stacked source images to enable photometric stereo processing is the most challenging element within the process.

An impression of the seal of Aršāma from Sigill. Aram. V.
An impression of the seal of Aršāma from Sigill. Aram. V.

The final recordings are incredibly impressive. Every tiny detail of the impression, historic repair and even the fingerprints of the maker are clearly visible.  These features can be explored using a 3D viewer within GIS software.  Moving over the surface of the recording is similar to flying over the surface of a desert landscape, where each granular element becomes a geographical feature. This new method of recording represents an important advance in imaging for the purposes of preservation.  The recordings of the seals will allow researchers to study originals in a way that has never before been possible.

In the left-hand example below, the shaded representation of the recorded surface has been generated by positioning a virtual light source at 60 degrees from the surface on which the original rests.  In addition, other shaders can be applied, as shown in the right-hand example, which uses a spectrum of colour to represent height.

A different perspective. Two renders of the surface of Sigill. Aram. VIII made with data recorded with from the Selene. Left: a greyscale shaded render. Right: a heat map, using a spectrum of colour to represent variations in height.
A different perspective. Two renders of the surface of Sigill. Aram. VIII made with data recorded with from the Selene. Left: a greyscale shaded render. Right: a heat map, using a spectrum of colour to represent variations in height.

Recording the seals in this way has made it possible to capture them at over six and a half million pixels per square inch, but at this resolution the depth of field is extremely shallow.  Focus stacking is a technique whereby multiple images are photographed from a static position with an incremental adjustment made to the focus between exposures.  The resulting stacks of images are then combined in software. In this way the depth-of-field is extended and the recording appears absolutely sharp from top to bottom.  Perfect alignment of the four focus-stacked source images to enable photometric stereo processing is the most challenging element within the process.

Combining focus-stacking and photometric stereo. Though the thickness of the seal is a mere 7.5mm, limited depth-of-field due to recording at such a high magnification only allows for acceptably sharp capture of the top 2mm. The benefits of focus stacking are particularly notable at the edges of the seal as they taper down. Left: single exposure. Right: focus-stacked image. Sigill. Aram. V.
Combining focus-stacking and photometric stereo. Though the thickness of the seal is a mere 7.5mm, limited depth-of-field due to recording at such a high magnification only allows for acceptably sharp capture of the top 2mm. The benefits of focus stacking are particularly notable at the edges of the seal as they taper down. Left: single exposure. Right: focus-stacked image. Sigill. Aram. V.

Every tiny detail of the impression, historic repair and even the fingerprints of the maker are clearly visible.  These features can be explored using a 3D viewer within GIS software.  Moving over the surface of the recording is similar to flying over the surface of a desert landscape, where each granular element becomes a geographical feature. This new method of recording represents an important advance in imaging for the purposes of preservation.  The recordings of the seals will allow researchers to study originals in a way that has never before been possible.

3D views of the reverse of Sigill. Aram. VIII. The wonderfully preserved string from this letter bulla still holds a fragment of parchment from one of the letters to which it was originally attached.
3D views of the reverse of Sigill. Aram. VIII. The wonderfully preserved string from this letter bulla still holds a fragment of parchment from one of the letters to which it was originally attached.

The image below shows one of the fourteen parchment letters from the Aršāma collection.  The Aramaic text is reasonably well preserved, and has been almost fully transcribed.  The letter suggests that Aršāma valued not only horses, two of which feature on his seal, but also three-dimensional artworks.  Addressed to Nakhthor, the steward of his estates in Egypt, Aršāma commissions the production of statues to be made by a sculptor believed to be Hinzani.

Ancient Persian letter (fragmented). Bodleian Library Pell. Aram. III.
A letter addressed by Aršāma, Persian Satrap of Egypt to Nakhthor the steward of his estates in Egypt. An excerpt of the text is translated as follows. …‘And let him make statues (on) which there shall be horsemen (?), and let him make a statue of a horse with its rider, just as previously he made before me, and other statues. And send (them), and let them bring (them) to me at once, with haste’… Pell. Aram. III.

So it seems fitting that we should carry out Aršāma’s request, albeit two and a half millennia later.  Producing a scaled-up three-dimensional facsimile of the fifth seal using the data recorded with ARCHiOx technology.  Firstly, the Lucida scanner was used to record the general shape of the seal from each orientation.  This volumetric data provided a base, over which the higher resolution, higher frequency data recorded with the Selene could be overlaid.

With the photometric stereo and laser recordings combined, elevated printing was then used to construct the facsimiles at four times the original size.  Several variations were made in order to assess which might be most useful for the purposes of study.  Firstly, an uncoloured version was made, showing only the volume of the seal. Two coloured versions followed, the first printed with a shaded render in order to enhance the debossed design, and the second printed with the albedo (colour) image recorded from the original seal.

Left: Two, scaled-up, 3D printed facsimiles of Sigill. Aram. V, made in the print rooms at Factum Arte, Madrid. Right: The two tiny facsimiles in the centre of the group are printed at actual size. Variations of enlarged facsimiles were produced, either uncoloured or with renders printed on their surface.

A far greater challenge would be to create a facsimile of the lost cylindrical seal which was used to make the impressions in the seven bullae.  Though the fifth, seventh and eighth seals provide much of the design, some elements are clearly incomplete.  A collated line drawing from Christopher J. Tuplin and John Ma’s book, Aršāma and his World: The Bodleian Letters in Context reveals two important missing elements from the design.  In the drawing, the horse to the left of the soldier holding a spear appears complete. Crucially so too does the inscription above the horse.  With the assistance of Professor Tuplin, these additional details were explained. Another seal bearing a partial impression, made using the same cylinder is held in the collections of the Persepolis Fortification Archive in Chicago.  A photograph of this seal was used by Eduardo Lopez from Factum Arte in order to incorporate the missing elements into the digital reconstruction.

The lost cylindrical seal, remade. The design from the collated recordings 3D printed onto flexible plastic before being glued to a cylindrical base. An impression in plasticine demonstrates that the facsimile is capable of creating incredibly similar designs to those found on the original bullae.

Prior to producing the facsimile, the 3D recording was inverted so that the embossed design would be capable of creating an impression similar to those from the original bullae.  Though limited by the resolution of the 3D printer, the facsimile cylindrical seal is indeed a usable tool and capable of making impressions which look very similar to those which were ordered to be made by Prince Aršāma, two and a half thousand years ago.

Download the full essay by John Barrett, Senior Photographer and ARCHiOx Technical Lead (Bodleian Libraries)

Logo of the Helen Hamlyn Trust

ARCHiOx, part 3: Patterns and paintings in a 17th-century Ragamala album

An essay by John Barrett, Senior Photographer, Bodleian Libraries, about discoveries from the ARCHiOx imaging project, which has been funded by the generous support of the Helen Hamlyn Trust. See also:  ARCHiOx: research and development in imaging – The Conveyor

An album of Ragamala paintings at the Bodleian Library (Bodleian MS. Laud Or. 149) is a beautifully painted manuscript, dating from the early 17th century. Not long after it was produced, the volume was donated to the Bodleian by Archbishop William Laud, at some point between 1635-41.

It has been proposed that that three recently discovered paper pouncing patterns may have been used in the production of paintings in the manuscript. The patterns, which have subsequently been loaned to the Bodleian, are skilfully made.  Tiny pin-pricks form the outline of illustrations which are clearly comparable with three of the paintings from the Ragamala Album.

Left: a paper pouncing pattern, photographed conventionally. Centre: an edited version of the previous image showing the position of the tiny pinholes. Right: A detail from fol. 8 of the Laud Ragamala Album. MS. Laud Or. 149.

Pouncing is a less obvious method of copying than pricking. Charcoal dust would have been transferred though the holes, duplicating the form of a design from pattern to page. Whether or not the three pouncing patterns were indeed the source of the paintings from the Bodleian’s 17th century volume remains somewhat of a mystery. In order to examine how closely the two align, the ARCHiOx team generated a set of renders from 3D recordings of the pouncing patterns and overlaid these with the colour images from the manuscript.

A layered image comprising of: Left: a painted page from the Laud Ragamala Album. Right: a mirrored heat-map render of the verso of the corresponding pouncing pattern. Centre: a composite of the left and right images. MS. Laud Or. 149.

Though some elements within the designs differ, there is a clear and extremely close correlation between the patterns and paintings.  3D imaging of the paintings themselves show no evidence of holes or depressions due to tracing, only the layers of pigment which have been applied to the paper.  Though the 3D recordings have not provided a definitive answer as to whether the patterns may be the origin of the paintings, it is hoped that they may serve as a template for similar analysis.

Download the full essay by John Barrett, Senior Photographer and ARCHiOx Technical Lead (Bodleian Libraries)

Logo of the Helen Hamlyn Trust

ARCHiOx, part 2: Digital imaging within a tradition of facsimile-making

ARCHiOx is by no means the first technology to create facsimiles of ancient texts or images. The process of copying using pinholes is evident on the largest original which has so far been captured for the ARCHiOx project. Dating to the 14th century, the Gough Map is one of the earliest maps to show Great Britain in a geographically recognisable form and served as a blueprint for maps of Britain for over 150 years.

Oblique images of the sign marking the location of Hull, East Yorkshire. Left: albedo. Right: shaded render showing the micro topography of this area of the map, in the absence of the original’s colour. Tiny indentations marking the form of the sign provide evidence that the map was copied from a precursor map. MS. Gough Gen. Top. 16.

Bequeathed to the Bodleian Library by Richard Gough (1735-1809), the map is covered in over two-thousand tiny indentations which transferred the position and form of geographical features from a precursor map.  Through studying these pinholes, researchers may be able to determine which features would have been present on the precursor map and in doing so, estimate when it may have been made.

This historic map has been recorded numerous times since its creation.  It therefore serves as wonderful case-study in the development of copying and imaging techniques.  A copper printing plate was engraved in 1780, prints from which are held in the Bodleian’s collections. Using a novel reproduction method developed at the Ordnance Survey, a photozincography recording was made in 1871.  In 1958, a run of collotype prints of the Gough Map were made at Oxford University Press.   The map was recorded digitally for the first time in 2006.  Hyperspectral and 3D laser recordings followed nine years later, in 2015.   These initial 3D recordings were conducted by the Factum Foundation’s Head of 3D scanning, Carlos Bayod.

“The recording carried out in 2015 applied the Lucida 3D Scanner to capture for the first time the topographical characteristics of this unique map. One of the first collaborations between the Bodleian Libraries and Factum Foundation, this survey allowed us to see and measure the shape and surface of the map without the colour layer, making it much easier to allocate the distribution of the pinholes, among other marks present on the relief. The information captured by the Lucida systems offers the possibility of visualizing the map’s surface on-screen as a shaded render, an image format onto which it is possible to register other layers of information such as the colour photographs. Additionally, it creates a greyscale depth map that can be used for re-materializing the data as an accurate physical reconstruction, becoming the base for creating an exact facsimile”. Carlos Bayod Lucini, Head of 3D Scanning, Factum Foundation

The new photometric stereo recording of the Gough Map captured with the Selene, was captured in June, 2022. MS. Gough Gen. Top. 16.

The photometric stereo captures made for ARCHiOx are the highest resolution recordings of the Gough Map to date.  Both the front and reverse of the map were recorded at over 700,000 pixels per square inch.  In order to record the map at this resolution, 85 image tiles were captured, processed and stitched together to form a single image.  Prominent pinholes and scoring marks are clearly visible from the recordings. These have been analysed, using geographical information system software by Damien Bove, Researcher for The Gough Map Project and Picture Editor of Imago Mundi: International Journal for the History of Cartography

“The pricking on the Gough Map is key to its creation, marking the location and form of place signs copied through from a precursor map. Where the tool has been pressed through the skin, it has left holes. Most of these can be seen on high resolution photos and on the earlier Lucida scan. Where the tool was pressed with less force, however, it has left only small depressions. The ARCHiOx scan has allowed us to identify and measure these for the first time, giving us a fuller understanding of the earlier map.” Damien Bove, Researcher for The Gough Map Project and Picture Editor of Imago Mundi: International Journal for the History of Cartography.

Visitors examine a three-dimensional facsimile of the Gough Map, made by Factum Arte, following a presentation given by the Bodleian’s Map Curator, Nick Millea.

But the ARCHiOx recording has not only allowed for on-screen analysis.  The data has also been used to create a remarkably accurate three-dimensional facsimile of the map.  Currently installed in the Bodleian’s Map Room, the facsimile provides an opportunity for close examination, ensuring that the original map need not be as frequently transported or removed from its protective casing.

“Facsimiles allow us to have a more natural connection with valuable cultural objects. Thanks to the possibility of reproducing the surface relief and colour in high resolution, a facsimile can serve a triple function contributing to the preservation, study, and dissemination of the original, for the benefit of both experts and amateurs alike”. Carlos Bayod Lucini, Head of 3D Scanning, Factum Foundation

— An essay by John Barrett, Senior Photographer, Bodleian Libraries, about discoveries from the ARCHiOx imaging project, which has been funded by the generous support of the Helen Hamlyn Trust. See also:  ARCHiOx: research and development in imaging – The Conveyor

Download the full essay by John Barrett, Senior Photographer and ARCHiOx Technical Lead (Bodleian Libraries)

Logo of the Helen Hamlyn Trust

ARCHiOx, part 1: Finding stories in the margins

An essay by John Barrett, Senior Photographer, Bodleian Libraries, about discoveries from the ARCHiOx imaging project, which has been funded by the generous support of the Helen Hamlyn Trust. See also:  ARCHiOx: research and development in imaging – The Conveyor

 

A 9th century insular manuscript, Gregory the Great, Homiliae XL in evangelia. MS. Laud Misc. 429.

The above manuscript, Gregory the Great, Homiliae XL in evangelia, is written in Latin and dates to the first half of the 9th century.  The 15th century shelfmark on folio 2, reveals that this volume was in the possession of the cathedral church of St. Kilian in Würzburg.  Examples of annotations made not in ink, but through scratching the surface of the parchment using a drypoint stylus have now been discovered and recorded on twenty-five pages from this volume, using the Selene.  The catalogue description for the recto of folio 74 shown in the image above, describes a drawing in the lower margin. A hunting scene, barely visible from the conventional photographic recording, but clear enough to make a partial digital annotation.  Far more successful at revealing the inscription, the 3D render shows not only the illustration, but also four camouflaged letters, R, O, D, A. This demonstrates how 3D recording can compliment traditional imaging in revealing and documenting new discoveries.

A shaded render of a drypoint addition from the lower margin of folio 74r.
A compiled digital annotation using conventional and 3D recordings, showing the position and form of the addition. MS. Laud Misc. 429.

The drypoint annotations recorded on folio 60r, in the image below, are inconsistent with the majority of others from this manuscript.  These have been added between passages of text rather than confined to the margins.  In this example, relatively deep incisions have been made, marking the position of punctuation. Far less obvious and perhaps only recognisable from the 3D render is a small, marginal illustration showing two hands, tied together with a bow.

A digital annotation from folio 60r, showing numerous drypoint additions. MS. Laud Misc. 429.

In order to determine whether or how this annotation might relate to the text, the image above was shared with Jo Story, Professor of Early Medieval History, Leicester University.  Her interpretation reveals a clear link between annotation and text.   The text from this homily describes the stoning of Stephen. The translation of folio 60r begins ‘when Stephen was dying for his faith, Saul kept the clothes of the stoners. Therefore, he himself stoned them all with his own hands, who returned all the works to the stoners.’  The connection between inscription and text is most evident from the passage at the end of the fourth line ‘Duo ergo sunt que’ – ‘because many are called but few are chosen’ – Chapter 22:14 from the Gospel of Matthew.  This passage immediately follows the verse ‘Then said the king to the servants, Bind him hand and foot, and take him away, and cast him into outer darkness, there shall be weeping and gnashing of teeth.’

Dozens of similar recordings of unlinked manuscript annotations have now been captured using ARCHiOx technology.  The discovery of the name ‘Eadburg’ from another of the Bodleian’s early medieval manuscripts by PhD candidate Jessica Hodgkinson (University of Leicester) is described in a previous Conveyor post.  Recordings from these two manuscripts have demonstrated that photometric stereo recording is extremely effective and is likely to hold the key to documenting incised markings from similar volumes.  Revealing these markings which have remained undetected for centuries is an incredibly exciting application of this new technology.

“The new photometric stereo recording methods that are being pioneered by John and the ARCHiOX team are transformative. The method allows us to see the surface of the pages in much greater detail than ever before and will give us insights into the preparation of the membrane and the methods used to make the quires, as well as acts of reading and engagement with the book after it was completed. New, and almost invisible, marks are now easily seen – revealing huge amounts of new information about medieval book culture – and the people who made and read them. This changes what we can do, the questions we can ask, and the answers that are revealed.” Jo Story, Professor of Early Medieval History, Leicester University.

Download the full essay by John Barrett, Senior Photographer and ARCHiOx Technical Lead (Bodleian Libraries)

Logo of the Helen Hamlyn Trust

Women in the Margins: Eadburg and Bodleian Library, MS. Selden Supra 30

by Jessica Hodgkinson and John Barrett

A series of exciting inscriptions, almost invisible to the naked eye, have been discovered in the margins of an important eighth-century manuscript in the Bodleian Library, MS Selden Supra 30. Revealed through state-of-the-art 3D recording technology by the ARCHiOx project, these marginal annotations provide tantalising new insights into this manuscript’s history and its links to women, in particular, to a woman called Eadburg.

Introducing Bodleian Library, MS Selden Supra 30

Bodleian Library, MS. Selden Supra 30 is a copy of the Acts of the Apostles, a book of the New Testament, written in Latin. It is a small volume, measuring only 229 x 176 mm (only slightly bigger than an A5 piece of paper).

Bodleian MS. Selden Supra 30 open at pp. 18-19
Bodleian MS. Selden Supra 30 open at pp. 18-19

Like most surviving manuscripts from this period, MS. Selden Supra 30 does not contain a formal colophon or scribal note recording when, where, and by whom it was made.

However, certain features of this manuscript, including the style of uncial script used to copy the text, demonstrate that it was produced in England, most likely somewhere in the kingdom of Kent, probably in the first half of the eighth century (i.e., between c. 700 and c. 750 AD).

MS. Selden Supra 30 was certainly in Kent by the fourteenth century when a shelf mark was added to p. 1 showing that it was then in the library of the monastery of St Augustine’s in Canterbury.

The opening page of MS. Selden Supra 30 (p. 1) which has suffered damage and is now significantly discoloured. The shelf mark from St Augustine’s, Canterbury (Di. I. G. III) is visible in the upper margin. 
The opening page of MS. Selden Supra 30 (p. 1) which has suffered damage and is now significantly discoloured. The shelf mark from St Augustine’s, Canterbury (Di. I. G. III) is visible in the upper margin.

Prayers added to p. 70, which was originally left blank, suggest, however, that very early in its history MS. Selden Supra 30 was owned and used by a woman.

These prayers were copied in the same type of script as the rest of the manuscript but by a different scribe to the two responsible for copying its main text.

The first prayer is a petition to God made by an anonymous woman, described as God’s “unworthy servant” (indignam famulam).

This strongly suggests that, at the time the prayer was added, MS. Selden Supra 30 was being used by a woman, or a group of women. The prayer may have been copied into the manuscript by a female scribe.

The formula of this prayer is unique and does not survive in any other manuscript. It could have been composed by the petitioner herself.

In 1935, in the first edition of Vol. 2 of Codices Latini Antiquiores, Elias Avery Lowe, then a Reader in Palaeography at the University of Oxford, suggested that another addition made to MS. Selden Supra 30 could provide further evidence of its links to women.

Lowe recorded, for the first time in print, that the letters EADB and +E+ had been incised into the lower margin of p. 47. He noticed that the letters had been cut into the parchment with force, apparently using a knife, slicing through the upper surface of the membrane.

ARCHiOx recordings of the letters inscribed at the bottom of p. 47.
ARCHiOx recordings of the letters inscribed at the bottom of p. 47.

Lowe suggested that these letters were abbreviated forms of the female name Eadburh/Eadburg.

Discovering Eadburg

Studying MS. Selden Supra 30 in the Weston Library’s Rare Books and Manuscripts Reading Room in 2022, Jessica Hodgkinson, a PhD student at the University of Leicester, funded by the AHRC Midlands4Cities consortium, spotted another inscription in the lower margin of p. 18. This inscription had never been noticed before. It was very small and almost invisible to the naked eye but appeared to contain Eadburg’s name written in full.

State-of-the-art technology has now not only confirmed this new inscription, but revealed several other instances of Eadburg’s name, alongside many more early marginal additions, incised into the parchment of MS Selden Supra 30. These discoveries provide new and exciting insights into the use of this book by a woman called Eadburg in eighth-century England.

Recording the inscriptions
by John Barrett

Scratched markings on the surface of a page are usually photographed using a single light positioned at a low angle. This simple principle is termed raking light.  However, through recordings made for ARCHiOx, it has been demonstrated that scratched markings may be far more effectively recorded using a technique called photometric stereo.

A description of ARCHiOx and an explanation of the technology and processes in use can be read in this blogpost:  ARCHiOx: research and development in imaging – The Conveyor

 The photometric stereo workflow adopted for ARCHiOx uses 2D images to record and store 3D information. These images map the direction and height of the original’s surface, and are processed into renders showing only the relief of the original with the tone and colour removed.

Renders produced using a photometric stereo workflow are superior to raked light images in three ways:

A 3D render lacks the excessive contrast of a raked light image making markings easier to discern. Through the use of software, it is possible to re-light renders virtually, giving complete control over the intensity of the shadow and highlight over the recorded relief of the original.

The ability to filter for different textural frequencies makes it possible to separate the scratched markings from the texture of material on which the markings have been made.

Renders can be re-lit virtually from any direction or height making it possible to reveal markings made along any angle.

In addition, the depth of a marking can be measured by examining a cross-section through it. The profile may also provide clues regarding the mark-making tool, in this case a drypoint stylus.

A cross-section through the lower horizontal line of the letter ‘E’. The depth of the drypoint inscription through this line measures around 18 microns (0.018mm).
A cross-section through the lower horizontal line of the letter ‘E’. The depth of the drypoint inscription through this line measures around 18 microns (0.018mm).

A photometric stereo recording of the near-invisible inscription on p. 18 was captured in May 2022.

The Selene, a prototype imaging system designed and built by the Factum Foundation, project partner for ARCHiOx, was used for the recording.

Multiple images were captured from the inscription before being processed, filtered, and enhanced. The resulting high-resolution shaded render shows only the three-dimensional surface of the page.  Through this new image, the drypoint inscription has been recorded successfully for the first time.

Top to bottom: An unedited render of the 3D surface of the bottom of page 18 with the inscription now visible.  An enhanced version made through tonal remapping.  A digitally annotated version.  The digital annotation with the render removed.

Top to bottom: An unedited render of the 3D surface of the bottom of page 18 with the inscription now visible.  An enhanced version made through tonal remapping.  A digitally annotated version.  The digital annotation with the render removed.

Subsequent analysis and processing, overseen by Jorge Cano, designer of the Selene, led to a new set of renders which enhance the markings further. These new images were created by compiling renders, re-lit virtually from multiple directions, and using a process called principal component analysis, or PCA.

The clearest recording of the drypoint inscription to-date. The inscription has been enhanced using techniques including virtual relighting, image stacking and principal component analysis.
The clearest recording of the drypoint inscription to date. The inscription has been enhanced using techniques including virtual relighting, image stacking and principal component analysis.

The lines which form this inscription are incredibly shallow. Even the most prominent are only 15-20 microns in depth, perhaps equivalent to less than a fifth of the width of a human hair. It is, therefore, unsurprising that the inscription is simply absent from a conventionally-lit colour image of this page.

Top: The drypoint inscription is absent from this capture recorded using conventional lighting.  Bottom: The digital annotation is applied at the exact position where it was recorded using photometric stereo technology.
Top: The drypoint inscription is absent from this capture recorded using conventional lighting.  Bottom: The digital annotation is applied at the exact position where it was recorded using photometric stereo technology.

Processing the data using computational methods has revealed an astonishing amount but analogue (i.e., human) intervention has still been required to digitally annotate the image to clarify the reading. Despite attempts to filter specifically for the inscription, shading from the texture of the parchment and its many tiny creases have proved almost impossible to remove. This makes it difficult in some areas, to rule-in or rule-out the presence of lines. An objective and cautious approach has been taken with the digital annotation. This has involved multiple imaging colleagues working independently to contribute to a set of annotations which could then be compared. Finally, the renders and digitally annotated images were shared with the researchers, allowing them to make their own observations and annotations with the benefit of context.

Subsequent recordings made for ARCHiOx have revealed that Eadburg’s name is spelled out in full five times on five different pages of MS. Selden Supra 30 (pp. 1, 2, 3, 12, and 18). On some of these pages, and elsewhere in the book, other abbreviated forms of this name, including E, EAD, or EADB, are also present.

Reading the inscriptions

The discovery of Eadburg’s full name etched several times into the manuscript’s margins definitively confirms Lowe’s theory that the letters previously identified on p. 47 are, indeed, abbreviations of the same name.

Eadburg’s name was copied out using letterforms common to all the newly identified inscriptions. The form of the A (an oblique line with an oval bow on the left) and the angular U and G are distinctive. This suggests that the same scribe may have made all of these additions. If so, it is at least possible that the scribe was Eadburg herself.

Readers and owners of early medieval manuscripts, both men and women, sometimes added their names to books, usually in ink, but occasionally, as here, in drypoint. Another early eighth-century example is the ink inscription that records, in Old English, that Abbess Cuthswitha owned a copy of Jerome’s commentary on the Old Testament Book of Ecclesiastes (now Würzburg, Universitätsbibliothek, M. p. th. q. 2, fol. 1r).

Eadburg’s name could also be a mark of ownership or evidence of reading. Although small in scale, and faint, someone, perhaps Eadburg herself, was evidently keen to preserve her name in the pages of this book to be seen by subsequent readers. What is unusual about Eadburg’s name, however, is that it appears here, in full or in abbreviated forms, 15 times.

Eadburg’s name is written on the opening page of the manuscript (p. 1). It overlaps part of the top of the enlarged decorated initial P which begins the text. Here, her name is preceded by a cross (+).

A series of ARCHiOx recordings of p. 1 of MS. Selden Supra 30:

Capture using conventional lighting (Bodleian MS. Selden Supra 30, p.1)
Capture using conventional lighting (Bodleian MS. Selden Supra 30, p.1)
Detail of the unedited 3D render of the inscription (MS. Selden Supra 30, p.1)
Detail of the unedited 3D render of the inscription (Bodleian MS. Selden Supra 30, p.1)
Digital annotation of the inscription (MS. Selden Supra 30, p.1)
Digital annotation of the inscription (Bodleian MS. Selden Supra 30, p.1)
Image showing the digital annotation applied at the exact position where it was recorded using photometric stereo technology (Bodleian MS. Selden Supra 30, p.1)
Image showing the digital annotation applied at the exact position where it was recorded using photometric stereo technology (Bodleian MS. Selden Supra 30, p.1)

The decision to etch the name over the top of the first letter of the text must have been deliberate. It establishes Eadburg’s presence in the book from the outset and connects her name intimately with the biblical text it contains.

On p. 2, her name is framed by a cartouche.

Detail of the recording of the lower margin of p. 2 showing the name Eadburg surrounded by a rectangular border.
Detail of the recording of the lower margin of p. 2 showing the name Eadburg surrounded by a rectangular border. (Bodleian MS. Selden Supra 30, p.2)

On p. 18, Eadburg’s name forms part of a multi-word inscription added to the lower margin. Here also, her name is preceded by a cross. Some of the following letters are easy to see, whilst others, especially those towards the end of the inscription, are difficult to make out, even with the benefit of the new visualisation techniques.

The most recent and clearest recording taken of the inscription, enhanced through virtual relighting, image stacking, and principal component analysis, appears to show, however, that, among the visible letters, there is a wynn (Ƿ), the Old English letter for W. This letter can be distinguished from the Rs in the inscription, including in the name Eadburg, by the form of the bow which is pointed and extends further down the vertical line of the letter than on R. The presence of a wynn shows that the inscription was written, not in Latin, but in the Old English vernacular language.

This inscription probably comprises three words. The name Eadburg is the subject of the statement, so we might reasonably expect the other letters to include a verb followed by the object.

A preliminary reading of the inscription is:

+ EaDBURG BIREð CǷ….N

+ Eadburg bears [cw….n]

Most of the letters in what appears to be the third and final word are unclear, with only CW– at the beginning and -N at the end remaining legible.

One Old English noun that could fill this position is cwærtern, meaning ‘prison’. Interestingly, the inscription is positioned beneath the beginning of the text of Acts 5:18 which describes the imprisonment of the Apostles by the high priest of the Temple and his followers because they had continued to preach the Gospel (…et injecerunt manus in Apostolos et posuerunt eos in custodia publica). If cwærtern is the third word in the inscription on p. 18, perhaps Eadburg sought to mirror the text, associating herself with the Apostles in their imprisonment.

Deciphering the drawings

Alongside Eadburg’s name, several intriguing drypoint drawings have also been discovered. Some are clearly human figures, though further investigation is needed to establish exactly who or what they depict. All the figures are very small. Several seem to have been made by incising a line around a thumb or finger to form the outline of the figure.

ARCHiOx recording of the lower margin of p. 9. Digital annotation applied in lower image showing two figures. The figure in the background has outstretched arms, and is reaching towards the figure at the front who appears to be holding up a hand to signal them to stop.
ARCHiOx recording of the lower margin of Bodleian MS. Selden Supra 30, p. 9. Digital annotation applied in lower image showing two figures. The figure in the background has outstretched arms, and is reaching towards the figure at the front who appears to be holding up a hand to signal them to stop.

The scene added to the lower margin of p. 11, which features at least three figures, may also include two E‘s. There appears to be an E, preceded by a cross, to the left of the first figure, and a second E, followed by a wynn (Ƿ) between the second and third figures. Could Eadburg have drawn this scene in drypoint and signed her work with her initial, as found elsewhere in the manuscript?

Recording of the dry-point addition in the lower margin of Bodleian MS. Selden Supra 30, p. 11.
Recording of the dry-point addition in the lower margin of Bodleian MS. Selden Supra 30, p. 11.

Identifying Eadburg?

Eadburg’s name or initials are etched into several pages, sometimes next to contemporaneous dry-point drawings. But who was she? More work on the newly discovered additions may bring us closer to answering this question.

We know of nine women called Eadburg living in England at some point between the seventh and tenth centuries (for details see the online Prosopography of Anglo-Saxon England). Other sources provide some tantalising clues that might help identify the Eadburg of MS. Selden Supra 30.

Charter evidence suggests that a woman called Eadburg was abbess of a female religious community at Minster-in-Thanet, in Kent from at least 733 until her death sometime between 748 and 761. As Lowe suggested in 1935, her dates and location correspond with the palaeographic assessment of the script of MS Selden Supra 30.

Abbess Eadburg of Minster-in-Thanet may also be the woman of the same name who corresponded with Boniface, the West Saxon missionary bishop and Church reformer. He became archbishop of Mainz in 732 and was martyred by pagans in Frisia in 754. Surviving letters show that Boniface held Eadburg in high esteem and that she sent books to him in Francia. He commissioned from her a deluxe copy of St Peter’s Epistles to be written in gold.

Boniface’s friend clearly had access to manuscripts and the means to make them. As such she is an especially strong candidate for the woman whose name was etched into the margins of MS. Selden Supra 30.

John Barrett is Bodleian Library’s Senior Photographer and ARCHiOx Technical Lead for the Bodleian.

Jessica Hodgkinson is a PhD candidate at the University of Leicester funded by the Midlands4Cities doctoral training partnership. Her research explores the participation of women in early medieval book culture in Western Europe through the analysis of surviving manuscripts commissioned, copied, owned and/or used by them.

 With special thanks to Jorge Cano, designer and engineer for Factum Arte and the Factum Foundation, for his work on enhancing the recording of p. 18, to Dr Philip A. Shaw, Teaching Fellow in the Department of English Studies at Durham University, for helping to decipher the Old English of this inscription, and to Professor Jo Story and Dr Erin T Dailey at the University of Leicester for their guidance and suggestions.

Logo of the Helen Hamlyn Trust Logo of the AHRC Logo of Midlands4Cities Doctoral Training Partnership

ARCHiOx: research and development in imaging

A photo-essay by the Bodleian’s Senior Photographer and ARCHiOx Technical Lead for the Bodleian, John Barrett  

From plate to print: Left to right, albedo, normal map, depth map, shaded render, composite and original print. A copper plate portrait of antiquarian, Anthony Wood, recorded as never before, captured using the latest in three-dimensional recording technology. Rawl. Copperplates e. 65.

Producing objectively accurate images from the books and manuscripts in the Bodleian’s collections to enrich our growing digital archive is the primary focus of the Imaging Services department.   On occasion, requests from curators or clients may require our photographers to use specialist imaging techniques such as recording originals using ultra-violet light or infrared cameras.  However, for the most part the aim is not to reveal hidden details, but to produce faithful digital reproductions.

In contrast, the aim of a new research and development project now underway in the Bodleian’s  Imaging Studio is to record items from the collections in three dimensions, using entirely new technology, in the expectation that discoveries will be made through recording surface detail at extremely high resolution.

Almost invisible when photographed conventionally, the fine etching on the reverse of Rawl. Copperplates g. 21 is revealed. This composite digital image combines a shaded render of the surface of the plate, layered with an albedo (colour) image. The image depicts Invidia (Envy). No extant print made using this side of the plate has yet been identified. The etched lines are extremely shallow, measuring 0.029mm in depth. The plate measures a little over 10cm.

ARCHiOx –Analysis and Recording of Cultural Heritage in Oxford–is a collaborative project, bringing together the Bodleian Libraries and the Factum Foundation.  Based in Madrid, the Factum Foundation specialise in high-resolution 3D imaging and have worked in cultural heritage institutions throughout the world, producing exceptional, three-dimensional facsimiles of artworks and artefacts.

During this one-year project, experts from the Factum Foundation will provide equipment and training in the use of new technologies and assist in exploring ways in which the output from the project can surfaced in Digital Bodleian.  As the project progresses it is hoped that through the collaboration between technicians at the Bodleian and the Factum Foundation, solutions to improve and streamline the technology will be identified.

With generous funding support from the Helen Hamlyn Trust, the ARCHiOx project will continue until January 2023.  Thereafter the technology will remain at the Bodleian making it possible to provide high-resolution 3D capture as an ongoing service.

Unlike perhaps more widely-known 3D capture techniques such as photogrammetry, the two state-of-the-art machines used for ARCHiOx use different principles for recording volume and are specifically designed for the capture of low-relief surface texture.  This makes them well suited to the recording of the primarily flat, but texturally rich originals from the Bodleian’s collections.  This high-resolution, low-relief capture has been termed ‘2.5D’ rather than 3D.

The 2.5D data produced during the project will serve two purposes.  Shaded renders make it possible to view the surface texture of an original while removing their visible tone and colour.  This allows for academic research from originals that contain textural details which are difficult to see and cannot be adequately recorded using traditional photographic techniques.  Alternatively, the data may be used to produce 3D facsimiles from items within our collections, allowing the material nature of the original to be reproduced.

An example of a shaded render. Without the colour and tone of the original, the shaded visualisation shows only the surface texture of the original. A section of an 18th-century copper printing plate, Rawl. Copperplates e. 59, featuring Archbishop William Laud.

The Selene is an entirely new solution for capturing 2.5D data and is being used for the first time in the Bodleian.  Using computational methods to extract very detailed information about the surface of an object, the Selene records multiple 2D source images, each captured with meticulously positioned lighting.   The Selene was designed by Factum Foundation engineer, Jorge Cano and uses a principle called photometric stereo.  Captures generated with the Selene during the last two months have proved that the technology is capable of recording surface relief at an incredible 25 microns, or 0.025mm.  This is over three times the resolution of any technology previously used to capture cultural heritage material by the Factum Foundation.  

The Selene Photometric Scanner capturing one of the Lister copper printing plates in the Bodleian Library.

Taking two weeks to build and refine, the Selene was installed by designer Jorge Cano and engineer Matt Marshall in the Bodleian’s Imaging studio, in early February.  The Selene uses a high resolution camera and four custom flash units, which together can be moved horizontally over the surface of the original.  Multiple customised electronic modules synchronise the movement of the motorised guides with the triggering of each sequence of flashes.  The Selene captures a series of image tiles at a resolution of 1040 pixels-per-inch.  The number of tiles, which is dependent on the size of the original can be programmed allowing for fast, automated capture.   

Man at a desk working on electronic components
Factum Foundation engineer and designer of the Selene, Jorge Cano, begins the assembly of the Selene Photometric Scanner in the Bodleian Imaging Studio.
Jorge Cano, testing and modifying the Selene Photometric Scanner, the only machine of its kind.

Unlike the laser recording system which has been used successfully for over a decade by the Factum Foundation, the Selene not only generates surface data, but can also capture colour.  This is hugely beneficial as not only does it make it unnecessary to capture the original twice but, due to being produced using the same source images, the shaded render showing surface texture and the colour image, known as the albedo, can be aligned perfectly to easily create a composite from the two.  We believe this composite image can be of great value to researchers, containing an exceptional level of detail and a real sense of the material nature of the original.   

A composite image, layering a shaded render of the texture with a shadowless colour image (albedo). This combination produces an image which shows every engraved line. (Lister Copperplates 858)

Created from multiple source images, each lit from different angles, the albedo is an exceptionally evenly illuminated and shadowless recording.  In some cases the albedo has a notable advantage over images produced using traditional lighting methods, which for 2D capture typically require just two primary light sources.  For originals with an uneven and highly reflective texture such as varnished paintings, creating an albedo may offer an effective solution for reducing unwanted highlights in reproductions. 

Factum Foundation engineer Matt Marshall installing custom flash units to the Selene.

Employing a very different principle to the Selene, the Lucida is a close-range, non-contact recording system that captures high-resolution surface texture data through the use of a laser and two tiny cameras.  This is a well-established solution for 2.5D capture, having been used by the Factum Foundation since 2011 during projects including the recording of the Tomb of Seti I, in Egypt.  The Lucida is capable of scanning taller originals, or those with greater vertical variation than the Selene and has been used extensively to produce data suitable for the manufacture of remarkably accurate 3D facsimiles.  Height data captured with the Lucida is incredibly accurate, though the resolution of the Lucida is significantly lower than the Selene in all three dimensions.  

The Factum Foundation’s Lucida expert, Carlos Bayod Lucini (right), describes the technology behind the Lucida 3D Scanner to Bodley’s Librarian Richard Ovenden (left) and PhD researcher Chiara Betti. Photograph: Nick Cistone

The Lucida was installed at the Bodleian’s imaging studio by Factum Foundation expert Carlos Bayod Lucini. The data generated from the Lucida has been vital in two ways.  Comparing data between the two technologies has made it possible to determine an accurate elevation factor, allowing height measurements to be correctly estimated and recorded for depth maps produced with the Selene.  Through combining the two data sets by overlaying the high frequency information generated with the Selene, and the more reliably recorded gentle, but taller gradients measured with the Lucida, it has also allowed for the production an incredibly accurate depth map, using the combined strengths of the two recording systems.  This technically ambitious process is an important achievement in 2.5D capture and will make it possible to create 3D reproductions at higher resolution and with more accuracy than previously achieved by the Factum Foundation. 

Factum Foundation technician Celeste Anstruther modifies and tests the Lucida 3D Scanner

A selection of 18th–century copper printing plates are amongst the first of the Bodleian’s originals to be captured with the Selene and Lucida.  Primarily from the Rawlinson collection, the plates include portraits of antiquarian Anthony Wood and 17th century Archbishop William Laud, as well as scenes, architecture and antiquities.  Plates from the Lister and Gough collections, the latter featuring portraits made from drawings attributed to William Blake, have also been recorded for the project.  Perfect for 2.5D capture using photometric stereo technology, copper printing plates have relatively flat surfaces and very shallow, highly detailed engraved lines.  The notable plates chosen for capture were selected by Co-ordinator of the Centre for the Study of the Book, Dr Alexandra Franklin and Chiara Betti, a PhD student at the University of London specialising in the research of the Bodleian copper plate collections on a Collaborative Doctoral Partnership, with advice from researchers who have worked on the Bodleian’s rare collections of copper plates and the associated publications.   As well as recording each plate’s text and illustrations, it is hoped that images produced during the project will reveal evidence of corrections, alterations, and degradation through use. 

A portrait of Edward III, from Gough Copperplates d. 107. The design is identified as the work of William Blake. These derivatives were processed using source images captured with the Selene. Left to right, shaded render, composite and albedo.

Notoriously difficult to capture using traditional photographic techniques, copper printing plates provide a number of challenges for the photographer hoping to record their surface.  There is little consistency in the material nature of the copper plates in our collections.  While some plates have been cleaned and are highly reflective with little change to their original colour and lustre, others, still bearing corrosion on their surfaces, reflect back virtually nothing and when reproduced, images may lack tonal variation and detail.

When photographing flat metal objects, the risk of capturing the reflections of the photographic equipment and the necessity to position the primary light source extremely close to the lens typically mean that a case-by-case approach is required.  This is a very different methodology when compared to the recording of paper and parchment originals, where a consistent workflow and continuous measurement is essential.

Using the Selene as an alternative to traditional photography has proven extremely successful, allowing us to record at a level of detail never previously achieved, and without having to navigate the complications previously associated with capturing metallic originals.

A tile image from Rawl. Copperplates e. 104. This composite was made by overlaying a shaded render and one of the colour, source images. The detailed engraving and colour would be extremely challenging, or perhaps impossible to reproduce using traditional photographic techniques.

Though capture of the source images using the Selene Photometric Scanner is relatively fast, currently the workflow required for processing the images is slow and reasonably complicated.  A number of software applications are required to generate the final derivative images, and one of the goals for ARCHiOx is to develop a more streamlined process.

Producing a normal map is the initial step in the process of creating useful derivatives such as shaded renders.  Normal maps are commonly used in CGI and computer game design.  Though the normal map is a 2D image, 3D information can be derived from the normal map because instead of simply recording a colour, each pixel represents a direction relative to the recorded surface of an original.   An entirely flat surface positioned parallel to the camera would be recorded on the normal map as a line, perpendicular to the original.  As the angle of the surface of the original changes, so too do the angles of the recorded lines, known as normal vectors.  Recording these normal vectors pixel-by-pixel makes it possible to map the surface of the original.  The direction of the normal vectors are defined by each pixel’s red, green and blue content.  Given that a copper printing plate is almost flat, the normal map represents the surface with an almost uniform purple colour.

A normal map produced with the images captured using the Selene. The starting point for producing useful derivatives, such as shaded renders.

Using the normal vectors from the normal map, a depth map can be generated.  This two-dimensional greyscale image uses tonal range to store elevation values.  It is processed at 16bits which allows for far more increments between tones to be recorded than in a standard 8bit image.  Through applying a Gaussian blur to the depth map, a derivative which records the gentle gradients over a wider area of an original can also be made.  In the workflow which has been established for the project, it is from the depth map that shaded renders can be created, using mapping software.   

Normal map to depth map integration. The angles of the normal vectors transform the flat profile of the normal map into a three dimensional surface. Diagram: Jorge Cano

In ordinary use, a geographic information system, or GIS application, can be used to create topographic maps and 3D visualisations of landscapes using aerial imagery.  By greatly increasing the scale factor, the same software can be used to map the tiny variations captured with the Selene and Lucida. The light direction and intensity can be configured, processing the depth map in to a highly detailed shaded view of the surface of the original. This shaded render can then be exported as a 2D image.

Though shaded renders provide an exceptional visualisation of the texture of an original, allowing researchers to virtually relight shaded renders for themselves is extremely useful as changes in the direction and height of the light can reveal details which may be hidden when recorded in a single shaded image.   In the case of the copper plates, engraved lines will either appear darker or lighter depending on the direction and height of the light.  As well as developing viewers capable of displaying and merging image layers produced during the project, Richard Allen, Andy Irving and Tim Dungate from the Bodleian Digital Library Systems and Services hope to develop tools that will permit this virtual relighting from the derivatives created from the 2.5D recording systems.

Virtually relit. Four snapshots of a real-time application written in Java. Using a normal map file as the source, it is possible for researchers to define the position of a virtual light source by simply moving the mouse pointer over the image. A section of Rawl. Copperplates e. 104.

One of the most important elements in developing the Selene is to establish an accurate elevation scale factor.  Height measurements are estimated when recording an original using the photometric stereo principle, and an elevation scale factor must first be assigned to the resultant depth map in order for accurate measurements to be made from it.  Comparing data from the Selene against data generated with a high-accuracy measuring device like the Lucida, or from an optical profilometer, has been essential in determining the correct scale factor. 

John Barrett (left) and Jorge Cano analyse a depth map using mapping software at the Factum Arte workshop, Madrid. The depth of a single engraved line can be determined by generating a cross section and measuring the vertical difference between peak and trough. Assigning an accurate elevation scale factor is essential prior to this analysis. Photograph: Matt Marshall

In order to reveal details from originals with extremely shallow relief, it has been useful to increase the scale factor and in doing so, exaggerate differences in relative height.  For instance, it has only been possible to produce usable shaded renders from many of the mezzotint printing plates captured for the project by increasing the scale factor.   

An extremely difficult test for 2.5D capture. A small section of a beautiful mezzotint copper plate, Rawl. Copperplates c. 41. The Selene manages to record the incredibly fine surface texture, but the detail in the shaded render can only be discerned by exaggerating the elevation scale factor. Left to right, shaded render, albedo and composite.

Creating and sharing an archive of detailed shaded renders will no doubt be extremely useful for researchers, but given that the data recorded for ARCHiOx is truly three-dimensional, the exciting possibility of creating accurate 3D facsimiles from items within our collections is entirely feasible.

For over twenty years, the Bodleian have archived hundreds-of-thousands of digital images, captured from our collections.  Through the use of technologies like the Selene and Lucida, we now have the capability of reproducing items more accurately than ever, not just as a two-dimensional representations, but as tangible 3D recreations.

This next-level development in preservation is not only important for the conservation of the original. Faithfully reproduced, three-dimensional reproductions will allow students to have a less restrictive, more hands-on experience of some of the more delicate and difficult to access items in our collections.  3D facsimiles may also be used as a substitute for originals while temporarily unavailable due to being exhibited or undergoing conservation treatment.

Two renderings of the surface of the same copper plate in close-up, showing the engraved lines
Truly three-dimensional. Two 3D views of the recorded surface of Rawl. Copperplates e.65, generated using mapping software. The second example shows a layered view, using the albedo and a shaded render. The depth of each engraved line, measured at around 60 microns, can clearly be seen.

Recreating an accurate and functional printing plate using data captured with the new photometric system is a demanding test for the Selene, but even more so for the elevated printing technology used to create the 3D facsimile.  While the Selene is able to record over 1000 pixels for every linear inch of original, Factum Arte’s state-of-the-art large-format 3D printer is limited to around half of this resolution.  Though this resolution has proved to be entirely adequate for the elevated printing of reproductions of artworks, the incredibly fine and often geometrically complex engraved details of copper plates are much harder to reproduce.

Commissioned by antiquarian Richard Rawlinson, Rawl. Copperplates e.65 is a copy of an earlier printing plate and features a wonderful portrait of local Oxford antiquarian Anthony Wood, whose manuscript and book collections are held at the Bodleian.  A print from the new plate, engraved in 1709 by Michael Burghers, appears in Rawlinson’s own copy of his work, ‘The Life of Mr. Anthony a Wood’.

So, not for the first time, though perhaps for the first time in 313 years, a new copy of the Anthony Wood plate would be remade, inked and pulled through a printing press.   This time however, the copy would be made using the very latest digital technology, in Factum Arte’s print room in Madrid.

A facsimile plate being 3D printed in the Factum Arte workshop, Madrid.

In order for the facsimile to be durable enough to be pulled through the printing press, it was initially necessary for the plate to be backed to a copper sheet.  Having prepared and inserted a base, the 3D files were then uploaded to the elevated printer.  The recently installed custom Canon Arizona printer deposits multiple layers of resin as the print head repeatedly travels over the base.  These incredibly fine layers, measuring between just 2 and 4μm are hardened using ultra-violet light.  Not only can the printer create texture, it can also reproduce the original’s colour.  The process is time consuming, taking several minutes to build even the shallow relief of the printing plate. 

The facsimile plate is made using an elevated printing process. Layers of resin are hardened using ultra-violet light.

With the elevated printing complete, the final challenge would be to print from the facsimile in order that comparisons could be made to the original prints from the Bodleian’s collections.  Eager to produce the first prints, founder of the Factum Foundation, Adam Lowe used both a modern press and a replica Goya press to produce a preliminary batch of 2022 editions.  The prints are impressive, reproducing the incredibly fine cross hatching surrounding the central portrait.  This is an impressive achievement given that it is the first time that a printing plate has been produced using the new photometric stereo recording system.  It is likely that differences between the quality of the original prints and the reproductions can mostly be attributed to the limitations of the 3D printing technology rather than the 3D data generated by the Selene, but future developments in both technologies will no doubt lead to increased accuracy.

Founder of the Factum Foundation, Adam Lowe, prepares the facsimile plate for printing.
The inked facsimile.
Man operating a star-wheel rolling press to print an engraving.
The print is pulled through a replica Goya press.

Having now captured dozens of the Bodleian’s 18th-century copper printing plates for the ARCHiOx project, it has been a pleasure to show visitors to the Bodleian’s Imaging studio both the originals and the newly recorded digital renditions. The results from the Selene have generated a great deal of excitement from curators, conservators and researchers.  Coming at the same time as the doctoral research of Chiara Betti, and following publications by Anna Marie Roos, Jeremy Coote, and Mark Crosby, this project extends the library’s efforts to make these previously neglected relics of printing and book history accessible to researchers.

This technology has enormous potential for the capture of cultural heritage material and has greatly exceeded the expectations of all involved in the project.  But equal to the enthusiasm for the new technology and its output, visitors have universally expressed a greater appreciation of the skill and dedication of the engravers who made and printed from the original copper plates.  These items deserve to be recorded as perfectly as technology will allow.  In doing so these wonderful objects can be shared digitally for the research and enjoyment of everyone.

A composite of a Rembrandt portrait captured with the Selene for Agnews Gallery, London.

But recording a selection of the Bodleian’s copper printing plates is only the initial focus of ARCHiOx.    Now that the technology has been proven and refined, other collections which will benefit from 2.5D capture can be recorded.  From a Rembrandt portrait to a volume of Japanese Ukiyoe prints, and a mysterious collection of incised palm-leaf manuscripts, the Selene and Lucida will be used to reveal further exciting discoveries and record originals as never before.

Developments and output from the project will be recorded in a future post.

A Ukiyoe woodblock print from Nipponica 373. Albedo and shaded render.

Text and images (unless otherwise credited) by John Barrett, April 2022

With thanks for their assistance in writing this article:

Jorge Cano, designer and engineer for Factum Arte and the Factum Foundation.   Jorge is the designer of the Selene, has established the photometric stereo workflow used for ARCHiOx and has been responsible for my training with this exciting new technology.

Chiara Betti, researcher of the Rawlinson copper plate collection.  The working title of Chiara’s thesis is ‘The Rawlinson copper plates at the Bodleian Libraries’.  Readers with an interest in Chiara’s research are encouraged to contact her at chiara.betti@postgrad.sas.ac.uk. The research is funded by the AHRC through the Collaborative Doctoral Partnership. See: https://www.glam.ox.ac.uk/early-modern-copper-plates-bodleian-libraries

The incredible work of the Factum Foundation is documented on their website www.factumfoundation.org

This exciting project has been made possible through the generous funding support of the Helen Hamlyn Trust.

Bodleian manuscripts on the New Testament Virtual Manuscript Room

MS. Gr. bib. d. 6 (P) on NTVMR

We are delighted to announce that digital images of over a hundred key manuscripts for the New Testament in Greek at the Bodleian Library are now available through the New Testament Virtual Manuscript Room (NTVMR), hosted by the University of Münster. The research team at the Institute for New Testament Textual Research (INTF) has scanned and transcribed archival microfilms and historic photographs of these collection items.

The NTVMR is an online open collaborative research environment focusing on the textual criticism and research of Greek New Testament manuscripts. The Institute for New Testament Textual Research conceived this platform and continues its development.

There are over 5,600 known Greek New Testament manuscripts. Approximately ninety per cent of these have images available on the NTVMR. The platform was initially designed for editing critical editions of the Greek New Testament, in particular the Editio Critica Maior (ECM). It is open access, which means that anyone with an email address can create an account and begin customizing their own workspace and creating their own projects.

Although the NTVMR hosts high-resolution colour images from many institutions, most of its images are from black and white microfilm resulting from photography expeditions undertaken by INTF staff in the 1960s through 1980s. The new Digital Bodleian image licensing terms waive the former requirement to apply for permission to reproduce Bodleian imagery for non-commercial purposes. This allows the NTVMR to display these images without restrictions under the Creative Commons Attribution-NonCommercial (CC BY-NC) licence.

About 122 manuscripts of the New Testament in Greek, or portions thereof, are housed at the Bodleian, many of which have microfilm available in the NTVMR. Digital Orientalist has published an overview of NTVMR. For questions about the NTVMR, contact Greg Paulson.

Open-access Bodleian publications on HathiTrust

Proposal for a suburban annex in ‘The Future of the Bodleian’: ‘Built of ferro-concrete, the building would be a striking, and perhaps not unsightly, feature in the Upper River landscape, and have a beauty like that of the grain-elevators at Fort William and Port Arthur on the Great Lakes in Canada.’

Bodleian Library Publishing are delighted to announce the release of 548 scans of historical publications from the Bodleian and other University of Oxford libraries on HathiTrust. HathiTrust is a partnership of university libraries to preserve digital scans of printed collections. Our release makes over 250 collection guides, catalogues, exhibition books, and histories to which the Bodleian Libraries are rights holders open-access throughout the world.

Highlights of the Bodleian collection

The Bodleian publications now available to read on HathiTrust are especially helpful for researching the manuscripts, archives, and rare books that now form the library’s special collections. While most of our catalogues are now accessed primarily in digital forms, many readers overlook introductory printed guides. You can now read The Bodleian Library: A Subject Guide to the Collections (2004) for an overview of the system, or R.W. Hunt’s introduction to the library’s intricate system of shelfmarks in the first volume of the Summary Catalogue (1953). For recreation, you might simply browse photos of the Bodleian in the early 1950s.

‘Rough sketch for a Book-store between the Bodleian and the Clarendon Building’, from ‘The Future of the Bodleian’

Bodleian history

Medieval and Byzantine manuscripts

Modern manuscripts and archives

Oriental manuscripts and rare books

Rare books

Maps

Commonwealth and African archives and manuscripts

We are grateful to everyone who made this project possible: Andrew Dunning, Samuel Fanous, Sarah Barkla, Ruth Mallalieu, Beth Gibbs, and Martin Kauffmann. We also thank our colleagues at Oxford University Press, who printed our books before the formation of Bodleian Library Publishing and supported this project.

Where did you get that hat?

from Alexandra Franklin (Bodleian) & Giles Bergel (Engineering)

re-blogged and updated from the Bodleian Ballads blog, 2012

2021: Updating the discovery of detachable woodblock hats in Bodleian broadside ballads, we’ve had a printing block made with the detachable hat, and a matching hat for the woman pictured on the same ballad sheet, ‘Unconstant Phillis’.

Newly-made printing blocks from a 17th-century ballad in the Bodleian Libraries, 4o Rawl. 566(121), showing a man and woman with hats
Newly-made printing blocks reproduced from images on a 17th-century broadside ballad in the Bodleian Libraries, 4o Rawl. 566(121)

The idea of applying image-matching software to the illustrations within the broadside ballad collection occurred to Giles Bergel in 2011. Under a John Fell Foundation-funded pilot project in that year Dr Bergel, in partnership with the Bodleian Library, asked Professor Andrew Zisserman of the Visual Geometry Group at Oxford to test their pioneering technology on a sample of broadside ballads.

The hat — in “Unconstant Phillis”

The hat — in “The Noble Gallant”

The first step was to obtain high-quality images. A sample of 800 ballads dating from the seventeenth century was chosen: this was the period when ballads were most typically illustrated, with a combination of commissioned and stock images. Photography, carried-out by the Bodleian’s Imaging Services studio, was funded with a grant from the John Fell Fund. High-quality colour images (600 DPI, 24-bit) were the result: these are now mounted within the current Ballads database and replace some of the older, lower-quality bitonal images which the Bodleian is hoping to fully replace.
Relja Arandjelovic at the Visual Geometry Group processed the new images and built a test site to demonstrate the image matching. The results were impressive: the pilot proved that software can match woodcut images on multiple printings, in varying conditions. The software, still under development, will be migrated to a new Bodleian Broadside Ballads interface under the current JISC-funded project.

What does “image match” offer to researchers?

Researchers working with ballads quickly notice that the same or similar woodblock-printed illustrations appear on multiple broadsides. The same hand-holding couple appears time and again on ballads of love; the same ships decorate songs of naval battles; the same cityscape appears, surprisingly, in ballads about London and Troy. And sometimes the same illustrations appeared in early modern English books and pamphlets, too. (See A guide to English illustrated books, 1536-1603, by Ruth Samson Luborsky and Elizabeth Ingram).

The repetition of generic illustrations in this way is often derided as evidence of ballads’ lowly status, but (as well as providing evidence for popular iconography), for early-modern bibliographic detectives, the re-use of woodblocks provides evidence for the date of a ballad’s printing: deterioration of  a woodblock might be a way to establish a timeline of broadsides that share the same illustration.

But there are other questions to be answered, sometimes questions we didn’t even know we needed to ask, which could be answered by selection portions of woodcut printed images: like, ‘Where DID you get that hat?’