Recovering Shared Heritage via Spectral Imaging: Problems, Solutions, Interpretations

1. Introduction

This session is planned as a complement to a proposed DH2016 tutorial that will demonstrate multispectral imaging, also organized by the Lazarus Project. The Lazarus Project, based at the University of Mississippi, makes multispectral imaging available to cultural institutions at low cost.

From time to time multispectral imaging garners attention in the press for its success in recovering text and images from damaged manuscripts, maps, and printed books. Yet even among digital humanities professionals, aspects of the process of making multispectral images, from the selection of candidate objects for imaging, to practicalities of setting up for multispectral imaging, the collection of the data, the methods to combine the images to recover the feature(s) of interest, and their interpretation and display, remain obscure. The planned demonstration of multispectral imaging will allow conference attendees to become familiar with the practical aspects of the process, particularly the equipment used, its configuration, and its functioning. The talks in this session will provide a view from a higher vantage point of the practices and results of multispectral imaging. Topics to be addressed include political situations that generate an urgent need for spectral imaging projects, challenges and solutions in the organization of large imaging projects, the different purposes for which spectral images can be used, best practices in image collection, the development of new image processing algorithms to recover text from spectral images, and the use and interpretation of spectral images by scholars.

The talks will be unified by an emphasis on solutions to specific problems in spectral imaging, and it is hoped that this emphasis will facilitate projects by other spectral imaging teams, and will also help scholars in identifying manuscripts and maps that are good candidates for textual recovery through spectral imaging.

The ninety-minute session will consist of six brief papers, as follows:

2. From Chartres to Timbuktu: Spectral Imaging in a Time of Crisis

Gregory Heyworth

The past hundred years has witnessed greater devastation to the world’s collections of ancient manuscripts than any other period since the eleventh century and the First Crusade. Between continued war and climate change, the threat is only increasing. This talk will consider the role of spectral imaging as a response to the political and environmental upheaval. Specifically it will treat various techniques of recovery from spectral and volumetric imaging to x-ray fluorescence on important war-damaged collections in Europe, as well as the logistical challenges of recovery projects on collections in politically unstable areas of the world.

3. The Sinai Palimpsests Project: The Recovery of Erased Texts in the World’s Oldest Library

Michael B. Phelps

St. Catherine’s Monastery of the Sinai, Egypt, was built to approximately its form in the mid-6th century and today maintains the world’s oldest continually operating library. It holds 4,549 manuscript codices, among which are 160 known palimpsest manuscripts. The palimpsests preserve erased texts in 10 languages that date from the late 4th century to the 12th century. Only three of these 160 palimpsests have ever been the subjects of sustained scholarly study and published. Hence, the palimpsests of Sinai represent a largely unexplored source not only for new texts from antiquity but also for evidence for reconstructing the literary history of the Eastern Mediterranean.

A collaborative project of St. Catherine’s Monastery and the Early Manuscripts Electronic Library (EMEL) seeks to recover the erased layers of Sinai palimpsests and make them globally accessible to researchers. Participating scientists are using state-of-the-art spectral imaging and image processing to render the erased texts legible; 23 participating scholars, scattered from Portland to Beirut, are identifying and describing the erased texts based on the image data; and EMEL and the University of California Los Angeles (UCLA) are preparing to host the images and scholarly descriptions online in service to St. Catherine’s Monastery. The project represents arguably the most extensive application to date of spectral imaging to cultural heritage.

This talk will survey the methods, innovations, and discoveries of the project. The survey will explore two recurring themes of the project, one technical and the other historical. First, challenges in the implementation of large-scale spectral imaging projects in the cultural heritage arena; and second, the significance of palimpsestation in the transmission history of late antique literature and languages.

4. The Evolution of Spectral Imaging in the Study of Manuscripts

Adrian S. Wisnicki

The Livingstone Spectral Imaging Project (2010-) is now in its second phase. The project applies spectral imaging to study some of the most damaged surviving manuscripts of David Livingstone (1813-1873), the celebrated Victorian traveler, abolitionist, geographer, and missionary. The first phase of the project (2010-2013) targeted Livingstone's 1871 Field Diary, and sought to use spectral imaging to recover now faded and illegible text that Livingstone had written crosswise over the pages of a single newspaper. The current second phase (2013-2016) centers on Livingstone's 1870 Field Diary, which is of a much more fragmentary nature that the diary previously studied by the project and, in fact, is quite legible under natural light. As a result, in this second phase, the use of spectral imaging has shifted from recovering lost or invisible text to using the imaging technology to explore material aspects of the manuscript, such as after-the-fact-additions (sometimes in other hands) and elements of page topography that can reveal details of the manuscript's passage through time.

Broadly speaking, therefore, the Livingstone Spectral Imaging Project offers a window onto the evolving nature of using spectral imaging technology to study manuscripts during the last five-odd years. The work of the project itself has, in turn, galvanized the development of a larger, but related project, Livingstone Online. Through the collaborative, interdisciplinary nature of its methodologies, the Livingstone Spectral Imaging Project has also established a loosely affiliated network of specialists who have contributed to the development of the two project phases. This paper will reflect on the evolution of this project and give attention to the implications of this evolution, particularly in establishing a collaborative methodology that is empowered by its distributed, at times informal, international character.

5. Reducing a Challenging Multi-Spectral Imaging Task to Practice

Kenneth Boydston

The multispectral image data set of a cultural heritage object can have a number of uses by disparate parties, and can be in and of itself historically significant. When the object is large, largely degraded, and of significant historical value such as is the late fifteenth-century world map by Henricus Martellus at Yale, acquiring the images that will satisfy known and anticipated demands can be particularly challenging. Time constraints, budget constraints, technological limitations, logistics, handling considerations, risk assessment, personnel, and facilities are among the parameters that can affect the outcome and which should be included in project planning.

Of particular interest are the technological capabilities and limitations. The technology can impact other parameters, and other parameters can impact the technology. While technology will certainly impact the outcome of a multispectral imaging project, it is often the case that technological limitations drive the project—though this should not be the case. In this paper we will suggest that multispectral imaging projects should be driven primarily by the needs of the scholars and conservators who will use the data, and that by focusing on these needs, technological innovations can be created and appropriate technology deployed that will make the difference between mediocrity and success. Examples from the imaging of the Martellus Map demonstrate a few such innovations.

One innovation required by the combination of budget, image spatial resolution, value, size and weight of the object was a large, yet easily portable easel capable of supporting the map and precisely moving both up and down and left and right over a grid of locations. Spatial resolution required that the images be captured in an 11 x 5 grid of tiles and stitched together. Software and hardware protocols were developed to facilitate manual movements of the map. Lasers were integrated to track and maintain focus and camera alignment.

Innovative use of software (MegaVision’s PhotoShoot Multispectral Imaging Capture Software) enabled a capture configuration customized for the particular needs of the map. This configuration was planned in advance, and then modified on site as the nature and needs of the map were revealed in preliminary captured samples. The magnitude of the image data set, together with the complexity of processing software (such as PCA), required innovative methods for preparing the imagery for visual appreciation.

6. Image Processing Techniques for Spectral Images of Historical Objects

Roger L. Easton, Jr., Ph.D.

The history of image processing of manuscripts to enhance or recover erased or damaged text goes back more than 100 years. The first work likely was that of the German physicists Ernst Pringsheim and Otto Gradenwitz in Breslau, who reported in 1895 on the development of an analog method for enhancing the visibility of the undertext of palimpsests by combining pairs of photographic transparencies that were collected under different conditions and processed differently. Their technique was improved by Fr. Raphael Kögel in the 1910s; he used ultraviolet illumination to further enhance the visibility of the undertext in one of the two images, which improved the results. These analog methods were difficult to implement, requiring careful photographic processing and accurate optical alignment to obtain good results. The first use of digital imaging algorithms to recover text from manuscripts may have been by Dr. John F. Benton at the Jet Propulsion Laboratory in the 1970s, who used contrast enhancement and image sharpening tools in the NASA image processing toolkit.

Over the last 30+ years, the capabilities for collecting and processing digital spectral images have improved to the point where it is now possible to have a complete spectral collection and processing system in a single suitcase that may be checked on an airline. The collected images may be processed using a variety of image processing tools, many of which had been developed for use in environmental remote sensing applications and are therefore not available in “general-purpose” image manipulation tools, such as Adobe Photoshop® and the GNU Image Manipulation Program (“GIMP”). Some are available as plugins in ImageJ, but we mostly use the special-purpose package ENVI that is written for remote sensing applications. This talk considers the image processing algorithms that were applied or developed for use to recover text from a wide variety of historical objects.

The algorithms may be loosely divided into two classes: deterministic methods for rendering the image data in pseudocolor, and custom methods that are based on the spectral statistics of the specific leaf. Among the methods in the latter category are principal component analysis (PCA), independent component analysis (ICA), and the minimum noise fraction transform (MNF). In all cases, preprocessing and postprocessing tools are useful. The preprocessing is applied before the deterministic and custom methods and is often necessary to calibrate the images or compensate the image data for differential fading or other problems. The postprocessing tools are used to combine processed images to improve the rendering for scholarly reading.

Examples will be shown of successful processing applied by the author and/or by collaborating team members to a wide variety of manuscripts and other objects in institutions all over the world. Among these are the Archimedes Palimpsest, the Syriac-Galen Palimpsest, the palimpsests in the “New Finds” at St. Catherine's Monastery, the Gruskovà palimpsest in Vienna, and the world map by Henricus Martellus (c. 1491) at Yale.

7. New Light on Henricus Martellus’s World Map at Yale (c. 1491): Multispectral Imaging and Early Renaissance Cartography

Chet Van Duzer

One of the outstanding problems in the history of cartography in the last half century has been that presented by a large world map made by Henricus Martellus in about 1491, which surfaced in the late 1950s and soon came to reside in the Beinecke Library at Yale. The map has long been thought to be one of the most important of the fifteenth century; in particular, there was good evidence that it influenced the thinking of Columbus with regard to his proposed voyage west to reach Asia; it was thought to have influenced Martin Behaim’s globe of 1492 and Martin Waldseemüller’s famous world map of 1507. But the vast majority of the texts on the map were illegible due to fading and damage, and thus its exact place in Renaissance cartography was impossible to determine.

In this talk I will look at the results of a recent NEH-funded project to make multispectral images of Martellus’s map at Yale—at the scholarly use that can be made of the images. These images have rendered almost all of the previously illegible texts on the map legible, and thus have enabled a detailed comparison with Martin Waldseemüller’s world map of 1507. This comparison shows that Martellus’s map was the source of most of the long descriptive texts on Waldseemüller’s map. At the same time, it turns out that the later cartographer chose not to follow Martellus for many of the other details of his map, and thus the comparison generates insights into the workshop practices of an early sixteenth-century cartographer, a subject about which we have very little documentation. In effect the multispectral images enable us to watch Waldseemüller at work, choosing different sources for different categories of information on his map as part of the process of creating a new image of the world.