Imaging the Venetus A Manuscript for the Homer Multitext

panel / roundtable
Authorship
  1. 1. Ryan Baumann

    University of Kentucky

  2. 2. W. Brent Seales

    University of Kentucky

  3. 3. A. Ross Scaife

    University of Kentucky

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

In May, 2007, a collaborative group of librarians, scholars and
technologists under the aegis of the Homer Multitext project
(HMT) traveled to Venice, Italy to undertake the photography
and 3D scanning of the Venetus A manuscript at the Biblioteca
Nazionale Marciana. Manoscritto marciano Gr.Z.454(=822):
Homer I. Ilias, or Venetus A, is a 10th century Byzantine
manuscript, perhaps the most important surviving copy of
the Iliad, and the one on which modern editions are primarily
based. Its inclusion will have the utmost signifi cance for the
HMT edition. In addition to the Iliad text, Venetus A contains
numerous layers of commentary, called scholia, which serve to
describe particular aspects of the main text. The scholia are
written in very small script (mostly illegible in the only available
print facsimile (Comparetti 1901)), and their organization on
the page appears as a patchwork of tiny interlinking texts (see
fi gure 1). The scholia are what truly set this manuscript apart,
and their inclusion is of the greatest importance for the HMT.
For this reason, we included 3D scanning in our digitization
plan, with the aim to provide digital fl attening and restoration
that may help render the text (especially the scholia) more
legible than digital images on their own. In this presentation
we will describe how we obtained both digital photographs
and 3D information, and will show how we combined them to
create new views of this manuscript.
[Figure 1: This slice from Venetus A shows the main
text, along with several layers of scholia: interlinear,
marginal (both left and right), and intermarginal.]
This particular manuscript presents an almost perfect case
for the application of digital fl attening in the context of an
extremely important, highly visible work that is the basis
for substantial and ongoing scholarship in the classics. In the
thousand-plus years since its creation, the environment has
taken its toll on the substrate, the parchment upon which the
texts are written. Several different types of damage, including
staining, water damage, and gelitination (degradation of the
fi ber structure of the parchment) affect different areas of the
manuscript. Much more widespread, affecting each folio in the
manuscript and interfering with the reading of the text, is the
planarity distortion, or cockling, that permeates the manuscript.
Caused by moisture, cockling causes the folios to wrinkle
and buckle, rather than to lie fl at (see fi gure 2). This damage
has the potential to present diffi culties for the scholars by
obstructing the text, especially the scholia. Virtual fl attening
provides us with a method for fl attening the pages, making the
text easier to read, without physically forcing the parchment
and potentially damaging the manuscript.
[Figure 2: Cockling is visible along the fore edge of the Venetus A]
Digital Flattening
The procedure for digital fl attening involves obtaining a 3D
mesh of the object (a model of the shape of the folio), in
addition to the 2D photograph. The photographic texture
information can then be mapped on to the 3D mesh, and the
3D mesh can be deformed (or unwarped) to its original fl at
2D shape. As the associated texture is deformed along with
it, the original 2D texture can be recovered. For the digital
fl attening of the Venetus A, the project used the physical-based
modeling approaches established in (Brown 2001, Brown
2004). The largest previous published application of this digital
fl attening technique was to BL Cotton Otho B. x, a signifi cantly
more damaged 11th century manuscript consisting of 67 folios
(Kiernan 2002, Brown 2004). By contrast, Venetus A consists of
327 folios, a nearly fi ve-fold increase. Thus, a system was needed
which could apply the necessary algorithms for fl attening
without manual intervention, in an automated fashion. We also
required a reliable system for imaging the document, acquiring
3D scans, and obtaining calibration data.
Imaging: 2D and 3D
Digital photographs were taken using a Hasselblad H1 with
a Phase One P45 digital back, capable of a resolution of
5428x7230 (39.2 megapixels). The manuscript was held by a
custom cradle built by Manfred Mayer from the University of
Graz in Austria. The cradle enabled the photography team to
hold the book open at a fi xed angle and minimize stress on
the binding, and consisted of a motorized cradle and camera
gantry perpendicular to the page being imaged. The camera
was operated via FireWire, to reduce the need for manual
adjustments.
To acquire data for building the 3D mesh, we used a FARO
3D Laser ScanArm. Since we were unlikely to be able to use
a fl at surface near the cradle to mount the arm, we set the
arm on a tripod with a custom mount. The ScanArm consists
of a series of articulating joints, a point probe, and a laser line
probe device attachment. The system operates by calibrating
the point probe to the arm, such that for any given orientation
of joints, the arm can reliably report the probe’s position in
3D space. Another calibration procedure calibrates the point
probe to the laser line probe. The result is a non-contact 3D
data acquisition system which can obtain an accurate, highresolution
untextured point cloud.
Imaging Procedure
As the document required a minimum of 654 data sets to be
fully imaged (verso and recto of each folio), data acquisition
was split into a number of sessions. Each session typically
represented a 4 hour block of work imaging a sequence
of either recto or verso, with two sessions per day. Fifteen
sessions were required to image the entire manuscript. Due
to the size and fragility of the codex, it was moved as seldom
as possible. Thus, each session was bookended by a set of
calibration procedures before and after placing the manuscript
in the cradle. In order for the extrinsic calibration to be
accurate, neither the FARO nor the cameras could be moved
during a session.
The process for a typical page was thus:
• Position the page and color calibration strips, then turn on
the vacuum to suction the page down
• Open the barn door lights on both sides of the cradle
• Photograph the page, via USB control
• Close lights to minimize damage to the manuscript
• Scan the page with the laser line probe on the ScanArm
(due to the small width of the line, this last step usually took
2-5 minutes)
Periodic spot-checks of the calibration data were carried out
to ensure that good data was being obtained. Test fl attenings
were also performed to check that everything worked as
expected.
Relationship between 2D and 3D
As the 2D and 3D data acquisition systems are completely
separate (unlike the tightly coupled systems of (Brown 2001,
Brown 2004, Brown 2007)), a calibration procedure was
required to determine the correspondence between 3D points
in the arm and 2D points in the camera. By calibrating the
camera and ScanArm in relation to one another, we were able
to determine the location of any point in the three dimensional
space within view of the camera and to fl atten the folios within
that space. As we were unsure of the amount of control we
would have over the primary photography, we used our own
camera (a Canon EOS 20D, which was attached to the camera
gantry immediately beside the Hasselblad) to obtain a set of
control images for each page and calibration data for each
session. The Canon was set to manual focus and a fi xed focus
and zoom, in order to have intrinsic camera calibration which
would remain relatively stable. As the goal of the Hasselblad
photography was to obtain the highest quality photographs
of each page, it did not retain the same focal length for every
page and would instead be focus on the current page depth. As
a result, intrinsics (including focal length, distortion matrix, and
center of radial distortion) were acquired using the average
focus that the camera would be at during a session. Once all
this information was obtained – photographs, 3D information,
and calibration – we were able to map the photographs to
the 3D mesh and unwrap it to its 2D shape, helping to make
legible texts that may otherwise be diffi cult to read.
The publication of legible, high-resolution digital photographs
of the Venetus A manuscript is the fi rst step towards an
electronic edition of the text, scholia, and translation which will
constitute a resource with far-reaching implications. Students
and scholars of Greek literature, mythology, and palaeography
could study in minute detail the most important manuscript
of the Iliad. This codex has been seen by only a handful of
people since its discovery and publication by Villoison in 1788
(Villoison 1788). Although Comparetti published a facsimile
edition in 1901 (Comparetti 1901), very few copies of even
this edition are available, even in research libraries, and the
smallest writing in Comparetti’s facsimile is illegible. The Homer
MultiText, with its high-resolution digital images linked to full
text and translation, will make this famous manuscript freely
accessible to interested scholars (and others) worldwide.
Bibliography:
Baumann, R. Imaging and Flattening of the Venetus A. Master’s
Thesis, University of Kentucky, 2007.
Brown, M. S. and Seales, W. B. “Document Restoration Using
3D Shape: A General Deskewing Algorithm for Arbitrarily
Warped Documents.” ICCV 02 (2001): 367-375.
Brown, M. S. and Seales, W. B. “Image Restoration of
Arbitrarily Warped Documents.” IEEE Transactions on Pattern
Analysis and Machine Intelligence 26: 10 (2004): 1295-1306.
Brown, M. S., Mingxuan, S., Yang R., Yun L., and Seales, W. B.
“Restoring 2D Content from Distorted Documents.” IEEE
Transactions on Pattern Analysis and Machine Intelligence 29: 11
(2007): 1904-1916.
Comparetti, D. Homeri Ilias Cum Scholiis. Codex Venetus A,
Marcianus 454. Phototypice editus, Praefatus est Dominicus
Comparettix. Lugduni Batavorum: A. W. Sijthoff, 1901.
Kiernan, K., Seales, W. B., and Griffi oen, J. “The Reappearances
of St. Basil the Great in British Library MS Cotton Otho B.
x.” Computers and the Humanities 36: 1 (2002): 7-26.
Lin, Yun. Physically-based Digital Restoration Using Volumetric
Scanning. PhD Dissertation, University of Kentucky, 2007.
De Villoison, J. B. G., ed. Homeri Ilias ad veteris codicis Veneti
fi dem recensita. Venice, 1788.

Conference Info

Complete

ADHO - 2008

Hosted at University of Oulu

Oulu, Finland

June 25, 2008 - June 29, 2008

135 works by 231 authors indexed

Conference website: http://www.ekl.oulu.fi/dh2008/

Series: ADHO (3)

Organizers: ADHO

Tags
  • Keywords: None
  • Language: English
  • Topics: None