Digital Images and the Technical Study of Art
a paper presented in the session
Annual Meeting of the College
Thank you Ron for organizing such an intellectually challenging session and for inviting me to speak on a topic which you are so qualified to address yourself.
To address the subject of this session, we might think of three levels of technology used in the study of art. First, studying art with normal vision alone. Secondly, as Andrea Kirsh and Andrea Bayer have just described for us, studying art not only with normal vision but also through technical examination. In this paper, I attempt to explore a third level, to see what additional types of information and ideas are available only when technical study is combined with computer technology.
Many people are, quite understandably, fascinated with computers. To some extent, perhaps we all are. But for the purposes of art history computers are primarily a means to an end, and I therefore intend to restrict this paper to one question: "In what ways can computer technology improve the technical study of art?
We shall attempt to answer this question with as clear a head as possible. Indeed, a clear head is very much required, given the hype of the computer industry and press, combined with the seemingly irresistible attraction of world-famous art for demonstrating the possibilities of computer manipulation. But it requires knowledge of historical context and considerable scholarly care to use digital technology for art history and conservation. Even in supposedly reputable publications, these qualities are not always in evidence. In her 1995 article in Scientific American, titled "The Art Historian's Computer: Riddles posed by ancient works of art fall to historical analyses and electronic explorations," [photo - photo] computer artist Lillian Schwartz claims that by flipping Leonardo's so-called Self Portrait and scaling it to match the Mona Lisa, she has discovered that Leonardo used himself as a model. Along the left edge we see Leonardo's drawing, flipped left to right and scaled, morphing into the Mona Lisa. Only a little less inventive than last week's New Yorker cover [photo].
In this paper I hope we can be open to the dramatic advances of computer technology without being misled by unrestrained enthusiasm. This requires, above all, that we have a firm grounding in our disciplines and a long-term view of what we hope to accomplish in our research and teaching.
Attempting to keep a clear head then, in what ways, if any, can computer technology improve the technical examination of art?
Many of the ways are now common to nearly all fields of human activity. For example, we use computers to create digital databanks that can then be consulted when we choose, even from distant locations. Most of the types of technical examination presented in this session already make use of computer technology in various ways, but most do not require them. In many cases they are merely an efficient convenience: very convenient, but in most cases not necessary to carry out the examination. Because twenty minutes is twenty minutes, I have slashed all such uses from my paper and will concentrate on distinctive ways in which computer technology not only facilitates technical examination of art but significantly improves the results, especially new types of technical examination that cannot be done at all without computer technology.
To examine this question, it will help to distinguish between research uses of computer technology in individual conservation studios and laboratories and, on the other hand, teaching uses of computer technology which present the techniques and results of that research and share it on stand-alone museum kiosks or more broadly on institutional servers and the Internet. We shall be looking principally at research uses but, to some extent, also at the teaching uses that interplay with research.
Leonardo, Ginevra de Benci
Fortunately, digital imaging has been used in the study of Leonardo not only by an inventive computer artist but also by a careful and informed Leonardo scholar. In his book, Leonardo da Vinci: Origins of a Genius, published last year, David Alan Brown, Curator at the National Gallery of Art, describes his reconstruction of Leonardo's Ginevra dé Benci, carried out in conjunction with the National Gallery of Art's Department of Imaging and Visual Services. In a helpful two-column endnote, he includes a detailed description of the computer processes used in the research and reconstruction. Would that everyone would make it possible for readers to follow exactly how the computer manipulation of their images has been carried out. At least in scholarly publications, surely, we must find ways, as John Cupitt has done at the National Gallery, London, to record and preserve the image trail in the file for the document itself.
Ever since the turn of the century, scholars have generally accepted the proposal that this portrait [photo], now almost square, must have been cut down from a half-length portrait similar to others at the time; and that the missing portion must have included hands, as in this closely related example at the Met. [photo], often attributed to Lorenzo di Credi. Adopting this common proportion, Brown proposes that the Ginevra must have been cut down by 1/3.
Turning for a moment to the painting verso [photo - photo], we see that the wreath has clearly been cut at the bottom. Assuming that the juniper stem was centered, Brown concludes that the picture has also been cut slightly along the left (as we look at the back), and there are in fact saw marks along that edge. Based on these hypotheses, Brown and computer image specialist Alexi Bryant produced this reconstruction of the back [photo]. This was done with the computer, largely copying portions of the upper half in creating the lower half and cloning the background. A skillful artist could, I think, have done this by hand. No doubt it would have been a more tedious operation, but I doubt that the computer was necessary for this reconstruction.
[photo] For reconstructing the portrait, however, David Brown tells me that the reconstruction would not have been possible without computer use [photo]. 1/2 inch has been added, this time of course along the right side, and the portrait has been extended by 1/3 to correspond to the common proportions of a half-length portrait of this type. [photo] The hands are taken from the famous Windsor drawing, which Müller-Walde already suggested for the Ginevra dé Benci in 1889, and which many scholars have endorsed. (We might note that it is easy to read the hands as one above the other, but Leonardo has in fact drawn both pairs of hands together twice, detailing her left hand in the drawing below and her right hand in the drawing above.) Using the lower drawing with the right arm in a more horizontal position, Brown and Bryant copied the drawing of her right hand from the upper drawing and pasted it in the lower position, but various subtle adjustments were made. The drawing is not a cartoon but a preparatory study. Obviously other missing parts have been painted in, also with the computer.
David Brown tells me that the reconstruction was worked on over the course of a year, that many slight variations were tried, and that he was impressed with how artistic the process was: both the computer work of Alexi Bryant and his own trial and error procedure based on years of study of Leonardo and his circle. The slides I have just shown you were taken from his book. [photo - photo] He kindly provided these slides from the forthcoming 2nd edition, in which a thin white line has been added to indicate more clearly the distinction between the existing painting and its proposed additions.
In terms of the various types of art historical or conservation use; this is an example of the virtual reconstruction of a damaged or partly missing work of art, an increasingly common use on a trial basis for actual conservation and restoration treatments. In terms of the different degrees of importance of computer use, this is an example of a study in which digital imagery was just a convenience for reconstructing the back but, because of the complex trial and error procedure, a near necessity for the front.
Another type of computer use that simply could not have been accomplished without the computer is exemplified in the research of Pepe Karmel, for the Jackson Pollock exhibition that just closed at the Modern, described in his impressive catalogue essay. Working with the original film footage and negatives of Han Namuth's famous photographs of Pollock painting (many more than were published in the book), Karmel reconstructed early stages of two of the paintings underway in 1950, when Namuth took his photographs. Putting together frames from film stills could easily have been done without the computer, since Namuth remained in one place and simply panned as Pollock painted. [photo - photo] In this film still composite of a different 1950 painting, we see at the top a frame of the left end and below the next frame simply added, overlapped.
However, putting together Namuth's still photographs had seemed impossible, because Namuth constantly changed position with the still camera, taking the photographs from a bewildering variety of angels and distances [photo - photo]. However, working with computer technician Mary Lynne Williams of Oven Digital in New York, Karmel began digitizing the angled views of Pollock's painting [photo] and, using the standard imaging software Adobe Photoshop, adjusting them to fit their shape and position on the finished canvas. Of course, some of this could have been done photographically. We regularly tilt the back of viewfinder cameras or the paper on which we are printing to adjust perspective, though rarely as much as this. What actually required the computer was adjusting the perspective on a number of photographs of the same painting, taken from different angles and distances, and combing them to achieve a partial view of the overall appearance of an early stage of the painting. The two paintings Namuth photographed in process in 1950, Autumn Rhythm at the Met. [photo] and One at the Modern [photo], are each over 17 feet long, and posed a major challenge for computer synthesis. As Karmel writes in his essay it was a "complex, time-consuming, and mind-numbing task." [photo] We see here a double page spread of Autumn Rhythm from the catalogue. Of course the existence of the finished painting, which we see again at the lower-right, provided a necessary template on which to superimposing photographs of earlier stages of the painting taken from different angles. At the upper left we see a very early state of the painting. Below that the completion of the initial configuration for the entire canvas. At the upper-right we see a later state, after it has been reworked in white, tan, and black; and the finished painting below.
Although one must be careful not to assume that this procedure was consistent throughout Pollock's career, looking at these images does allow Karmel to reach several important conclusions which correct earlier interpretations of his work. [photo] Karmel observes that early configurations of Autumn Rhythm correspond closely to Pollock's previous stick figures and suggest that he was not thinking in completely abstract terms as he began such paintings. [photo] We are looking here at the two intermediate stages, in slides directly from the digital files, which Pepe kindly provided. Perhaps most importantly, Karmel observes that at first Pollock created a series of largely self-contained configurations, and that the overall painting, which has been the subject of such critical debate, developed only during the intermediate stages as he elaborated and gradually obscured the initial configuration.
Pepe was kind enough to walk me through his Pollock research step by step and pointed out how similar the approach was to his doctoral dissertation on early Picasso drawings, where he used photographic overlays photocopied onto plastic sheets. One finds again and again that the computer has not inspired whole new approaches to research but that it facilitates and improves approaches already established within each discipline.
Shroud of Turin
Lawrence Doyle, Jean Lorre, and Eric Doyle
Developed by NASA for studying the surface of planets from spacecraft, computer intensive image processing techniques were first applied to works of art (or at least to an historical artifact) in the highly publicized study of the Shroud of Turin [photo]. This article was published in 1986 in Studies in Conservation, by Lawrence Doyle and Jean Lorre, of the NASA Jet Propulsion Laboratory, and by Eric Doyle. In the Shroud of Turin study, a digital image of the shroud [photo] was manipulated in various ways to eliminate as many as possible of the accidental features in the digital image (such as creases in the cloth, soil spots, scorch marks from fire, and the cloth weave) and thus to clarify the image of the head [photo]. Both of these slides show digital images, the one on the right with accidents removed. The important point here is that these were not removed individually by hand, with a computer stylus, but rather using mathematical programs that identified different types of inconsistencies and irregularities and then eliminated them in global commands. Also using global commands, details of the head were brought out to make the head as legible as possible [photo - photo]. Again the slide on the right shows the accidental features removed. All such techniques, now familiar to computer scientists, involve identifying the digital signature of one or another feature, in order to separate and eliminate that feature, wherever it appears, from the digital image. In the Shroud of Turin study, one of the features removed from the final image was the weave of the cloth. Of course, if one wanted to study the weave itself, one would remove the image of the head in order to clarify the weave [photo].
One of the most complicated such manipulations from the Shroud of Turin study could be applicable to the technical study of many works of art. In the Shroud itself there are color differences resulting from the materials used in the creation of the Shroud and in the creation of the head (by whatever means), also color differences from pollution, soiling and scorching over the years. But these color differences are so slight that all the human eye can see is slight shades of grey. With the computer one can eliminate much of the grey and exaggerate the slight but remaining color differences [photo], so that the colors are distinguished from each other and identifiable. This allowed researchers to see what marks were caused by scorching, what marks by blood, and what by other types of soiling. (These slides were loaned by Iowa State University and by Jean Lorre, member of the NASA Jet Propulsion Laboratory's Astronomy Image Processing Group.)
It's useful to understand the principle involved here since it lies behind many kinds of computer use that are beyond the ability of normal human perception. The example is sometimes given of a vast bowl of marble, each comparable to one pixel. The marbles are all blue, except that some have an extremely slight tinge of green. The human eye, evolved to perceive the full visible spectrum, cannot distinguish two colors that are so nearly identical. Normally, the computer can't either. But one can set the computer to perceive only the blue and blue-green portion of the spectrum, eliminating everything else. With the computer so narrowly focused, the difference between the two colors is greatly magnified and the marbles with a slight tinge of green are readily identified.
Investigating the Renaissance
Research in the Laboratory
Infra-red Reflectogram Assemblies
In the next two examples, we see how computer technology sometimes makes possible a significant improvement in the legibility of technical images, and, in the second example, how the computer sometimes helps us to discover relationships among images. Neither of these examples would have been possible without computer technology.
The first example involves the assembly of infra-red reflectograms in studying underdrawing. Because the television systems used for infra-red reflectography have relatively low resolution, it is necessary to take a large number of small, close-up reflectograms to record the underdrawing sharply and in detail [photo - photo]. Previous to computer use, photographs were taken directly from the video screen and then laboriously cut and assembled by hand, as in this assembly of the feet of St. Francis, from Van Eyck's St. Francis Receiving the Stigmata in Turin. This assembly was made in 1985 and is here shown from the 1997 book published by the Philadelphia Museum of Art, a splendid comparative study with their smaller, nearly identical painting. The entire Turin painting [photo] is only 11 by 13 inches so we are looking at a very small section of the assembly, beautifully done by the inventor of the technique and acknowledged master, Van Asperen de Boer. Yet what we are most aware of, even in this detail, is the mosaic of the parts, because individual reflectograms vary from light to dark across even such small surfaces. This makes the evidence more difficult to read and severely reduces the coherency of the drawing.
Infra-red reflectograms are now captured digitally and assembled on the computer. Although IRR assemblies were possible without computer use, this speeds the process and facilitates more precision in the assembly. More importantly, the computer allow us also to correct the tonal distortions of the reflectograms, so that each drawing, even if large, can be read as a coherent whole [photo - photo], as in this infra-red assembly of Wolf Huber's Christ Taking Leave of His Mother, from the 1997 National Gallery [London] Technical Bulletin. The painting is over three feet tall yet the reflectogram reads as a whole.
For those who may not be familiar with the procedure, the computer process is instructive. The technician-photographer takes a reflectogram of each little section of the painting and, in addition, using the same video set, takes a reflectogram of a standard, completely uniform, monochrome graycard. Since the reflectogram of the graycard has the same tonal distortions as the reflectograms of the painting (caused by the uneveness in infra-red video systems), one can subtract the distortion recorded on the reflectogram of the gray card from each reflectogram of the painting. Such images are superior for both evidence and presentation and would not be possible without computer technology.
An especially efficient and accessible procedure for this was developed in 1992-93 by Henry Lie, Director of the Conservation Center at the Fogg, and used by Ron Spronk in his research on the Fogg's early Netherlandish paintings. The results of that research were used in the actual cleaning and restoration of the paintings, which were then put on display in the Renaissance galleries at the Fogg.
I want, for the moment however, to switch from research uses of the computer in individual conservation labs to computer presentation of the research to others; because this is such an informative example of the multiple formats in which the results of research can be presented - using computer technology. In this case, the results were published by Ron in the Fall 1996 Bulletin of the Harvard Art Museums [photo], published and illustrated with computer technology, of course, like nearly all publication these days. The research was also presented in a stand-alone kiosk [photo] in the corner of one of the Renaissance galleries. On the right wall of the gallery [photo] we see the Jan Provoost Last Judgment, one of the three paintings discussed on the Kiosk display. Many of the images and text on the kiosk appear also in the Bulletin , so that one may reasonably ask what advantages, if any, the kiosk has, other than being attractive to computer addicted viewers. There are, in fact, a few clear advantages. The most under-appreciated advantage of digital images for art history and conservation is that one can make available on high quality computer monitors many more large, high quality images than can possibly be afforded in printed paper publication. This allows the Fogg kiosk to illustrate aspects of the research more extensively and in greater detail than was possible even in the well-illustrated issue of the Bulletin. This instructive kiosk was created by Ron Spronk and Robin Marlowe, and these slides were provided by Ron.
Unlike hardcopy publication, computer display allows for animation. The Fogg kiosk includes an animation of the infra-red reflectogram assemble procedure we have just discussed. One of the paintings which Ron studied in detail was the Portrait of a Man by the Master of the 1540's [photo - photo] which we see on the far wall in the gallery. On the computer kiosk, we see an image of the 6 reflectograms presented separately [photo]. Clicking a button, we are shown an animation [photo - photo - photo - photo] in which the 6 reflectograms come together, one at a time, and are trimmed to fit perfectly. To save time, I'm skipping the middle 8 images. Here are the last 4 and the final infrared assembly [photo - photo]. Of course, it would be possible in the catalogue to reproduce all sixteen of the images in this computer animation, but this number of images in a catalogue is impractical and more importantly doesn't convey the process as convincingly as the computer animation.
Even so, the computer animation is just a demonstration. We click the button and the reflectograms gradually come together and are trimmed as we watch. There is one section of the Fogg kiosk that stands out as an exemplar of the unique advantages of computer technology both in technical research and in presentation of that research. Using a procedure developed by Henry Lie, Ron was able to superimpose photographs of this same Portrait of a Man taken under different lighting conditions (normal light, infrared, x-ray, and ultraviolet light) which show different layers of the painting and thus different information about the painting. Adobe Photoshop allows a skilled technician to superimpose these images perfectly and to control each layer separately. Thus one can overlay all of the images and vary the intensity and transparency of each in whatever way and in whatever combinations one wishes to discover a multitude of otherwise invisible relationships. This is one of the most distinctive uses of digital imagery I have seen for the technical examination of works of art.
This layering of images of the Portrait of a Man was used not only in research on the painting, but was also presented as a demonstration on the Fogg kiosk and (of special importance) also on the Internet. [photo] We are looking here at an infra-red image of a detail of the Portrait of a Man (we could actually select other details if we wished). It is here overlayed with the same detail taken under natural light, though with this black tab at the extreme left we can see only the infra-red image. However, by grabbing this black tab [photo] we can gradually increase the percentage of the natural light image [photo]; until it takes over 100% at the far right [photo]. Even on the Internet, this is not an animation, so the viewer can move the tab back and forth at will, examining the painting at leisure and with care, to discover whatever relationships exist between these two layers of the painting. There is one image of this overlay in the catalogue [photo], but, after having used the web site, one longs to be able to slide the tab oneself, to be able to study the relationship between these two layers in whatever way one wishes.
I stress the importance of this more experimental use of the computer because it comes close to allowing the viewer to participate in the same type of research activity as experts in the lab. Making it possible for students, and the public, to engage in the kind of free exploration that most closely approximates research activity is far more productive educationally than straight-forward presentation, and I believe the same general principal of learning applies to the use of computer technology. I do not agree that the computer completely alters our ways of thinking. It is a tool that is gradually making it possible for us to teach (in schools and museums) in ways that many have been attempting to do all along and to reach vastly greater audiences in doing so.
There are a number of other types of computer use that we should look at and I apologize to colleagues who have shared their research with me but whose work I have not even mentioned. It pains me especially not to have time to look at examples of computer technology in the study of sculpture, most famously in studies of Michelangelo's Florentine Pieta. Here digital imaging now allows us to synthesize hundreds of photographs of a sculpture, taken from all angles, so that the sculpture becomes partly transparent, making it possible for us to see the relationship of forms at front and back or top and bottom. And, of course, we should also examine the now well-know ability of the computer to allow us to turn these three-dimensional models in whatever way we wish, in a process of free exploration. And, of course, we should look at computer uses in archaeological research, which in some areas have led the field.
The potential exists for an immense increase in the visual evidence for the man-made and natural worlds. It is important to emphasize that there is no correlation between degrees of technological complexity and content value. However, new types of technology can and do stimulate new ideas. Given the rapid development of computer technology itself, and even more our own primitive stage of experience in using it, it seems nearly certain that new avenues of research will be opened. Finally, one of the notable advantages of the computer revolution is the breakdown of traditional disciplinary barriers, furthering the long desired goal of a community of scholars.