Recent work in the EDUCE Project

Popular methods for cultural heritage digitization include
fl atbed scanning and high-resolution photography. The
current paradigm for the digitization and dissemination of
library collections is to create high-resolution digital images
as facsimiles of primary source materials. While these
approaches have provided new levels of accessibility to many
cultural artifacts, they usually assume that the object is more
or less fl at, or at least viewable in two dimensions. However,
this assumption is simply not true for many cultural heritage
objects.
For several years, researchers have been exploring digitization
of documents beyond 2D imaging. Three-dimensional surface
acquisition technology has been used to capture the shape
of non-planar texts and build 3D document models (Brown
2000, Landon 2006), where structured light techniques are
used to acquire 3D surface geometry and a high-resolution
still camera is used to capture a 2D texture image.
However, there are many documents that are impossible to
scan or photograph in the usual way. Take for example the
entirety of the British Library’s Cotton Collection, which was
damaged by a fi re in 1731 (Tite 1994, Prescott 1997, Seales
2000, Seales 2004). Following the fi re most of the manuscripts
in this collection suffered from both fi re and water damage,
and had to be physically dismantled and then painstakingly
reassembled. Another example is the collection of papyrus
scrolls that are in the Egyptian papyrus collection at the British
Museum (Smith 1987, Andrews 1990, Lapp 1997). Because of
damage to the outer shells of some of these scrolls the text
enclosed therein have never been read - unrolling the fragile
material would destroy them.
Certain objects can be physically repaired prior to digitization.
However, the fact is that many items are simply too fragile
to sustain physical restoration. As long as physical restoration
is the only option for opening such opaque documents,
scholarship has to sacrifi ce for preservation, or preservation
sacrifi ce for scholarship.
The famous Herculaneum collection is an outstanding
example of the scholarship/preservation dichotomy (Sider
2005). In the 1750s, the excavation of more than a thousand
scorched papyrus rolls from the Villa dei Papiri (the Villa of
the Papyri) in ancient Herculaneum caused great excitement
among contemporary scholars, for they held the possibility of the rediscovery of lost masterpieces by classical writers.
However, as a result of the eruption of Mt. Vesuvius in A.D. 79
that destroyed Pompeii and also buried nearby Herculaneum,
the papyrus rolls were charred severely and are now extremely
brittle, frustrating attempts to open them.
A number of approaches have been devised to physically open
the rolls, varying from mechanical and “caloric”, to chemical
(Sider 2005). Substrate breakages caused during the opening
create incomplete letters that appear in several separate
segments, which makes reading them very diffi cult. Some efforts
have been made to reconstruct the scrolls, including tasks such
as establishing the relative order of fragments and assigning to
them an absolute sequence (Janko 2003). In addition, multispectral
analysis and conventional image processing methods
have helped to reveal signifi cant previously unknown texts.
All in all, although these attempts have introduced some new
works to the canon, they have done so at the expense of the
physical objects holding the text.
Given the huge amount of labor and care it takes to physically
unroll the scrolls, together with the risk of destruction caused
by the unrolling, a technology capable of producing a readable
image of a rolled-up text without the need to physically open
it is an attractive concept. Virtual unrolling would offer an
obvious and substantial payoff.
In summary, there is a class of objects that are inaccessible
due to their physical construction. Many of these objects may
carry precise contents which will remain a mystery unless and
until they are opened. In most cases, physical restoration is
not an option because it is too risky, unpredictable, and labor
intensive.
This dilemma is well suited for advanced computer vision
techniques to provide a safe and effi cient solution. The EDUCE
project (Enhanced Digital Unwrapping for Conservation and
Exploration) is developing a general restoration approach
that enables access to those impenetrable objects without
the need to open them. The vision is to apply this work
ultimately to documents such as those described above, and
to allow complete analysis while enforcing continued physical
preservation.
Proof of Concept
With the assistance of curators from the Special Collections
Library at the University of Michigan, we were given access to
a manuscript from the 15th century that had been dismantled
and used in the binding of a printed book soon after its
creation. The manuscript is located in the spine of the binding,
and consists of seven or so layers that were stuck together,
as shown in fi gure 1. The handwritten text on the top layer is
recognizable from the book of Ecclesiastes. The two columns of
texts correspond to Eccl 2:4/2:5 (2:4 word 5 through 2:5 word
6) and Eccl 2:10 (word 10.5 through word 16). However, it was
not clear what writing appears on the inner layers, or whether
they contain any writing at all. We tested this manuscript using
methods that we had refi ned over a series of simulations and
real-world experiments, but this experiment was our fi rst on
a bona fi de primary source.
Figure 1: Spine from a binding made of
a fi fteenth-century manuscript.
Following procedures which will be discussed in more detail
in our presentation, we were able to bring out several layers
of text, including the text on the back of the top layer. Figure 2
shows the result generated by our method to reveal the back
side of the top layer which is glued inside and inaccessible. The
left and right columns were identifi ed as Eccl. 1:16 and 1:11
respectively.
To verify our fi ndings, conservation specialists at the
University of Michigan uncovered the back side of the top
layer by removing the fi rst layer from the rest of the strip of
manuscript. The process was painstaking, in order to minimize
damage, and it took an entire day. First, the strip was soaked in
water for a couple of hours to dissolve the glue and enhance
the fl exibility of the material which was fragile due to age; this
added a risk of the ink dissolving, although the duration and
water temperature were controlled to protect against this
happening. Then, the fi rst layer was carefully pulled apart from
the rest of the manuscript with tweezers. The process was
very slow to avoid tearing the material. Remaining residue was
scraped off gently.
The back side of the top layer is shown in fi gure 3. Most of
the Hebrew characters in the images are legible and align well
with those in the digital images of the manuscript. The middle
rows show better results than the rows on the edges. That is
because the edge areas were damaged in structure, torn and
abraded, and that degraded the quality of the restoration.
Figure 2: Generated result showing the back side of
the top layer, identifi ed as Eccl. 1:16 and 1:11. Figure 3: Photo of the back side of the top
layer, once removed from the binding.
Without applying the virtual unwrapping approach, the choices
would be either to preserve the manuscript with the hidden
text unknown or to destroy it to read it. In this case, we fi rst
read the text with non-invasive methods then disassembled
the artifact in order to confi rm our readings.
Presentation of Ongoing Work
In November 2007, representatives from the Sorbonne in
Paris will bring several unopened papyrus fragments to the
University of Kentucky to undergo testing following similar
procedures to those that resulted in the uncovering of the
Ecclesiastes text in the fi fteenth-century manuscript. And in
June 2008, a group from the University of Kentucky will be at
the British Museum scanning and virtually unrolling examples
from their collections of papyrus scrolls. This presentation at
Digital Humanities 2008 will serve not only as an overview
of the techniques that led to the successful test described
above, but will also be an extremely up-to-date report of the
most recent work of the project. We look forward to breaking
down the dichotomy between preservation and scholarship
for this particular class of delicate objects.
Bibliography
Andrews, C.A.R. Catalogue of Demotic Papyri in the British
Museum, IV: Ptolemaic Legal Texts from the Theban Area. London:
British Museum, 1990.
Brown, M. S. and Seales, W. B. “Beyond 2d images: Effective 3d
imaging for library materials.” In Proceedings of the 5th ACM
Conference on Digital Libraries (June 2000): 27-36.
Brown, M. S. and Seales, W. B. “Image Restoration of
Arbitrarily Warped Documents.” IEEE Transactions on Pattern
Analysis and Machine Intelligence, 26:10 (October 2004): 1295-
1306.
Janko, R. Philodemus On Poems Book One. Oxford University
Press, 2003.
Landon, G. V. and Seales, W. B. “Petroglyph digitization: enabling
cultural heritage scholarship.” Machine Vision and Applications,
17:6 (December 2006): 361-371.
Lapp, G. The Papyrus of Nu. Catalogue of Books of the Dead in
the British Museum, vol. I. London: British Museum, 1997.
Lin, Y. and Seales, W. B. “Opaque Document Imaging: Building
Images of Inaccessible Texts.” Proceedings of the Tenth IEEE
International Conference on Computer Vision (ICCV’05),
Prescott, A. “’Their Present Miserable State of Cremation’:
the Restoration of the Cotton Library.” Sir Robert Cotton
as Collector: Essays on an Early Stuart Courtier and His Legacy.
Editor C.J. Wright. London: British Library Publications, 1997.
391-454.
Seales, W. B. Griffoen, J., Kiernan, K, Yuan, C. J., Cantara, L.
“The digital atheneum: New technologies for restoring
and preserving old documents.” Computers in Libraries 20:2
(February 2000): 26-30.
Seales, W. B. and Lin, Y. “Digital restoration using volumetric
scanning.” In Proceedings of the Fourth ACM/IEEE-CS Joint
Conference on Digital Libraries (June 2004): 117-124.
Sider, D. The Library of the Villa dei Papiri at Herculaneum. Getty
Trust Publications: J. Paul Getty Museum, 2005.
Smith, M. Catalogue of Demotic Papyri in the British Museum,
III: the Mortuary Texts of Papyrus BM 10507. London: British
Museum, 1987.
Tite, C. G. The Manuscript Library of Sir Robert Cotton. London:
British Library, 1994.