Frequently, images of rare documents must be taken under strict time constraints, when a chance opportunity arises, and with equipment that is less than ideally suited for precise digitization. This can often result in uncalibrated images whose contents are nonetheless qualitatively useful. Due to the logistics of document imaging, perspective distortion is a common artifact which can manifest itself in images taken under these constrained circumstances. We propose an automated approach for correcting this perspective distortion, even for uncalibrated images of documents with irregular page edges containing no regular text.

In this paper we present an application of image registration techniques to the specific domain of manuscript images. We show the application of this technique to images of the Venetus A, a 10th century manuscript of Homer's Iliad. The same algorithm is used to register images of the MS across time (including photographs separated by over a century), as well as across imaging modalities.

Opaque artifacts containing obscured text continue to present significant challenges to conservators. Often, any attempt to read such fragile texts fundamentally and irreversibly alters the physical structure of the object in which they are contained. Thus, the conservator is presented with an almost impossible dilemma: risk destruction of an artifact in order to recover a text of unknown value, or maintain the artifact in its original state and possibly allow a valuable text to go undiscovered. Numerous types of fragile artifacts exist containing text which is unreadable under visible light from the object exterior. Papyrus rolls, in particular carbonized papyri such as those from Herculaneum and Tanis, present obvious difficulties along these lines. Given the huge amount of labor and care it takes to physically unroll these 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. 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 other types of closed documents, and to allow complete analysis while enforcing continued physical preservation. Our paper and presentation will present our successful experiments on a variety of objects, both synthetic and authentic. We have constructed a number of proxy scrolls which use papyrus and iron-gall ink, in order to perform initial testing of the concept. Additionally, we have conducted experiments on a 15th century manuscript of Ecclesiastes which had been reused in a book binding. In this case, although the composition of the ink was unknown, we were able to non-invasively read multiple layers of previously hidden text using a custom CT setup with a resolution of 30 microns. Conservators were then able to disassemble the binding layer-by-layer and confirm our results. We will also discuss future uses of this technique, both for near-term projects and potential long-term applications to specific types of artifacts. INTRODUCTION Opaque artifacts containing obscured text continue to present significant challenges to conservators. Attempts to read such fragile texts may fundamentally and irreversibly alter the physical structure of the object in which it is contained. Thus, the conservator is presented with an almost impossible dilemma: risk destruction of an artifact in order to recover a text of unknown value, or maintain the artifact in its original state and possibly allow a valuable text to go undiscovered. As a result, non-destructive, non-invasive techniques which can provide insight into such texts are a valuable topic of investigation. 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 enclosed text-bearing documents, and to allow complete analysis while enforcing continued physical preservation. In this paper, we describe the preliminary results of our experiments in imaging texts using micro-CT techniques, as well as discuss future applications of this technique and obstacles it faces. THE DAMAGE OF PHYSICAL UNROLLING Numerous types of fragile artifacts exist containing text which is unreadable under visible light from the object exterior. Papyrus rolls, in particular carbonized papyri such as those from Herculaneum and Tanis, present obvious difficulties for the conservator. Even papyrus rolls for which a conventional "unrolling" is feasible must be handled with care in preservation, as the length of a roll may cause it to be cut into sections and stored (and accessed) in this new format. There are also various examples of artifacts where a text's substrate has been reused in the creation of a new object, such as papyrus used in Egyptian mummy cartonnage (Wright 1983) or discarded parchment manuscript folios in the binding of a codex manuscript or printed book. The history of previous efforts to read and conserve such objects provides an important context for the advantages of a repeatable method of inspection which does not require physical intervention. Papyrus scrolls which can withstand unrolling often underwent the procedure soon after acquisition, being dampened and humidified to assist the process (Leach 1995). Carbonized papyri present more obvious frustrations for those wishing to read them. First, it is not possible to physically unroll carbonized scrolls completely. Following some years of experimentation (including practices such as pouring mercury through the edges of the scrolls, immersing entire scrolls in mercury; immersing scrolls in rose water; and holding scrolls in a chamber with various types of gases), common practice developed to cut away the hard outer layers the husks and unroll the relatively more flexible center layers (Sider 2005). The husks, consisting of two, four, or sometimes six separate pieces, must then be dealt with. Attempts to separate the layers often resulted in damaged or even completely destroyed text. Both chemical and physical methods were used, but as with the complete scrolls there was little success. There is still damage to come, even after scrolls are successfully opened and unrolled. Fading ink, color changes to papyrus, weakening structure of papyrus, and the oxidization of ink that causes papyrus breakdown are all documented damage that can come from exposing the inner layers of a papyrus scroll to light and air. In addition, after unrolling and the separation of layers, scrolls and scroll fragments were at times attached to a backing material. This introduces new problems for conservators and editors: texts written on both sides of the papyrus are either mounted in a frame with substantially less support for the center of the document or with one side of the document permanently covered, and the backing material, if acidic, can damage the object causing discoloration and weakening of the substrate. Some fragments are mounted between two planes of glass, but this exposes the papyrus to another set of physical stresses including the possibility that the glass will simply crush the parchment to powder (Leach 1995). There are also very specific editorial issues attached to the scrolls (Sider 2005). During the 19th century there was a program to document as many scrolls as possible. At this time practice was to read visible text and create a hand-drawn facsimile of it, then to scrape off the layer to uncover the text underneath, completely removing the top layer. For many scrolls, the only sources we have left are these hand-drawn facsimiles. Since the people who did this work were not scholars and did not know Greek, these documents must be seriously scrutinized by the editor. Another editorial issue involves the ordering of the scroll fragmentsonce cut into pieces and divided into layers, the original order of scroll texts was often lost. Editors are still dealing with this issue today. CT SCANNING AND VIRTUAL UNROLLING All of these examples serve to illustrate the underlying risks of destruction and harm to the physical artifact that traditional methods of autopsy entail and editorial problems that they introduce. The act of physical investigation frequently becomes a one-time endeavor, fundamentally altering the structure of the object and all future attempts at scholarship. In light of these limitations, non-invasive volumetric imaging techniques seem an ideal fit for the problem of analyzing opaque artifacts. The technique we explore here is high-resolution micro-CT imaging, which has proven a versatile and powerful tool in our initial experiments. Getting a usable result from X-ray based CT depends upon the specific X-ray absorption characteristics of the object, such as the substrate and pigments involved. However, as we will discuss, X-ray imaging gives an almost ideal response for a large variety of the inks we are interested in. Its high energy radiation is able to penetrate the full depth of most objects, and its physical configuration enables construction of portable imaging equipment suitable for performing analysis on-site. Most other volumetric imaging techniques do not have these advantages. They are either unsuitable to the nature of the materials, require a large fixed facility, or simply have not matured and commoditized to the point where constructing a portable device would be feasible. The ideal X-ray response is to have some signal from the substrate (to assist in segmentation and unwrapping, as well as to understand the overall physical configuration), and an increased response from the ink to provide contrast. The primary division among inks used for written texts is between carbon black and inks made with metallic or mineral content. Iron-gall inks, composed of galls, vitriol (iron sulfate), gum and water, are very prevalent in ancient writing. Due to the iron content of these inks, their X-ray attenuation is relatively high and ideal for our methods of investigation. In our testing of proxies which used ink with iron content, the contrast between text and papyrus is strong and consistent. One can anticipate similar X-ray response characteristics with other pigments which use metals or minerals to achieve their coloring. Aside from carbon black, this encompasses the majority of pigments used across a variety of cultures: red ochre (Fe2O3), azurite (2CuCO3*Cu(OH)2), verdigris (Cu(CH3COO)2*Cu(OH)2), malachite (CuCO3*Cu(OH)2), yellow ochre (Fe2O3*H2O), vermilion (HgS), orpiment (As2S3), pararealgar (As4S4), Egyptian Blue (CaCuSi4O10), Han Blue (BaCuSi4O10), Han Purple (BaCuSi2O6)

It has been shown in computer vision that combining reflective components (e.g. mirrors) with refractive components (e.g. lenses) can greatly increase the flexibility and functionality of an optic imaging system, such as increased field of view and reconstruction from a single image. Cameras with both lenses and mirrors are typically referred to as catadioptric systems. A projector, as the dual of a camera, can also benefit from catadioptric optics. For example, it is possible to bend the light rays to optically compensate for distortions caused by projection [Swaminathan et al. 2004]. However, different mirror shapes have to be used for different setups.