This leaflet has three purposes. It first defines digital technologies from the perspective of communication and coding. Then it describes the key components of a digital imaging system and the most important steps in the digital imaging process. Finally, it asks some big questions that ought to be considered as libraries and archives move from experimenting with the technology to using it as a tool for transforming the way they do business.
We are living in a digital world. The evidence is everywhere. Keyboards outnumber office workers. Everybody has a web page. Nobody carries cash. We are hearing words like "bitslag," "jitterati," "NIMQ," and "CGIJoe" in everyday conversation. Billionaire technologists seem to own all the digital copies of all the art that matters. There seems to be a growing concern in libraries and archives that if we are not going digital, being digital, or dreaming digital, then we are relegating ourselves to the great museum of paper.1
And yet, it may be that our biggest challenge may not be embracing digital technology but rather building a common language to describe the transformations that are having such a phenomenal impact on our everyday lives. A shared vocabulary is the key element in the development of a community of practice and a shared vision of the future among those of us who have responsibility to shepherd the nation's cultural resources. Jim Taylor and Watts Wacker note that "Looking backward, the true legacy of Naisbitt's Megatrends or Toffler's Third Wave may turn out to be not the worldviews but the words."2 Nowhere have the words mattered more than in our view of the place of preservation in the digital world in which we live.
At their most fundamental level, digital technologies are an extension of the long history of the way we communicate with each other. The desire to communicate provides the motive and the ultimate rationale for the development of technologies of all sorts. Today's digital world is concerned with creating, sharing, and using information in digital form. Digital information is data that are structured and manipulated, stored and networked, subsidized and sold.
Information takes many forms. One way to think about these forms is to distinguish between symbolic information and coded information. Let us illustrate this by looking at the many ways that the most common letter in the roman alphabet — the "E" — can be represented, beginning with the early symbols of the printed alphabet.
The period from the time Gutenberg invented in the middle of the fifteenth century through the year 1500 is referred to generally as the period of incunabula. During this time printers and book designers went to great lengths to make their products—type faces, format, and layout—look and function much like the manuscript books of the preceding centuries. Only when a theory of the alphabet and a theory of the book emerged around the time of Geofroy Tory's classic text on the structure of the roman alphabet were book designers able to begin taking full advantage of Gutenberg's technological innovation.3
Figure 1 is an illustration of the capital letter "E" from Tory's Champ Fleury of 1529, which sought to develop a theory of the alphabet around the proportions of the human body and the basic principles of Euclid. Here the letter "E" is a pattern of ink on paper.4
The world defined by strings of 1's and 0's has existed for a long time. The idea of the digital computer originated over 300 years ago in the fertile mind of a German mathematician, Gottfried Wilhelm von Leibnitz. In 1679, Leibnitz imagined a device in which binary numbers were represented by spherical pellets, circulating within a kind of pinball machine controlled by a rudimentary form of punched cards. He described a comprehensive numerical system in which all calculation can be expressed in combinations of 1 and 0—the identical approach that all digital technologies use today.5
We are living in an era of digital incunabula — a period marked by furtive efforts to make our digital products look and behave as their analog relatives do. Only when we have developed a theory of digital representation of information will we begin to take full advantage of Leibnitz's mathematical innovation. That theory is emerging today.6
Figure 2 is another symbolic pattern — Braille. Here the letter "e" is represented by large and small raised dots in a predictable grid. Note, too, that the same pattern can mean either the letter "E" or the number "5" depending upon the context in which the pattern is located. Context is another idea that is fundamental to the representation of information in digital form. With Braille, if you know the context and understand the pattern, communication is fast and efficient.
American Sign Language is symbol as signal. It is a language in which the form and motion of the hands combine to convey meaning. Form without motion is only half of the process. Communication depends upon a shared understanding of the meaning of both components of the language. Figure 3 is a static representation of the letter "E."
With semaphore, however, the pattern of motion is the symbol. The transformation from one formation of flags and arms to another establishes the communication link. Figure 4 is yet another static representation of the letter "E." Emerging theories of digital communication have yet to account fully for the multiple senses that we routinely use to communicate directly—the subtleties of body language, gesture, and inflection. As sophisticated as digital communication has become, its dependence on machines is seriously limiting.
Some of the earliest modern forms of direct communication over long distances, however, have been digital in character. Figure 5 is an illustration of Lord George Murray's Visual Telegraph that operated for a time from London to Deal beginning in 1794.7 The system consisted of raised platforms placed horizon to horizon. On each platform a large board had six large circular holes that could be closed by wooden shutters — strikingly familiar to the patterns of Braille — manipulated by a trained operator. Reports indicate that a message could reach along the chain of fifteen stations in a few minutes. But just think about the administrative overhead!
The route from the Visual Telegraph to modern digital communication is marked by successive transformations from symbol to code. Samuel F.B. Morse invented his digital code of dots and dashes as the language of his telegraph. The origins of radio — or wireless telegraphy — lie in the desire to extend the digital communication of Morse where wires could not reach. An early application of the analog technology of continuous waves was the transmittal of Morse's dots and dashes to ships at sea. The modern coding of the letter "E" as the ASCII code 01100101 owes its lineage to the theories of Leibnitz and the practical technology of Samuel Morse rather than to the technology of radio and television.
A digital system uses numbers to represent a concrete object or an abstract idea. Digitization is the process of transforming the object or idea into a numerical code. The baseline of digital technology is a coding system with only two numbers — 1 and 0 — hence the term binary. Each numerical place in the system is a bit. In the digital world bits are things; they take up space; they take time to move from one place to another. A collection of bits can be described and counted, much like anything.8 The most common way to count the bits in a system is by "byte" or eight bits, even though computer technology abandoned the byte as a discrete object decades ago.
A bit-mapped image is a digital picture made up of row after row of bits in a grid.9 In a digital image, a bit is commonly referred to as a pixel, short for "picture element." As objects, digital images are described in terms of three characteristics: resolution, dynamic range, and pixel size.
More recently, a fourth term, tonal value, has been applied to describe the characteristics of a "digital image," confusing terminology about a digital representation of an image, such as a photograph. A bitmap is a digitally coded pattern, not a digitally coded symbol such as text we recognize through an alphabet.
Resolution is the number of pixels (or dots) used to code a linear inch of surface horizontally and/or vertically. Visualize a piece of graph paper. The number of small blocks in a running inch up or down the paper is the resolution. The more pixels per inch the higher the resolution and the more accurately the patterns visible on a given surface can be represented digitally. The description of an image as 300 dots per inch (dpi) means that 300 pixels are used to represent each inch across the horizontal surface. It is sometimes (mistakenly) assumed that an image with 300 pixels horizontally will also be represented by 300 lines vertically. The actual structure of the digital grid depends on the capabilities of the scanning device.
Figure 6 is a 3 mm letter "e" at 600 dpi resolution scanned from negative microfilm at Yale University Library. Note that the digitally coded pattern occupies some 4,900 bits in the computer system compared to the eight bits required for the digitally coded symbolism of ASCII code.
Dynamic range refers to the number of possible colors or shades of gray that can be included in a particular image. Dynamic range is sometimes called "depth" and is commonly represented as bits per pixel. In bitonal scanning, the sampled image level for each pixel is rounded to 0 (black) or 1 (white). One bit of information is required to code the value of the pixel. In 8-bit gray scanning, the sampled image level for each pixel is rounded to one of 256 values, each representing successively lighter shades of gray. Eight bits of information are required to represent each pixel. In full-color scanning, the three hues of the color system are represented by one of 256 possible shades and encoded as a total of 24 bits (8 bits per hue). The two predominant color systems are Red/Green/Blue for monitor projection and Cyan/Magenta/Yellow for digital printing.
Pixel size is an important measure of the capability of a given piece of scanning hardware to represent the patterns of a surface completely. The"real resolution" of a scanner is the proportion of the surface that is detected. The"addressable resolution" of a scanner is number of pixels in a running inch of an array without optical correction. Greater real resolution depends upon the quality of the electrical and mechanical engineering of a given device. Scanner manufacturers sometimes use software solutions (synthetic resolution) to compensate for limited real resolution. It is important to be wary of scanner manufacturer claims and to undertake rigorous testing and benchmarking before committing to the purchase of scanning equipment.10
Tone reproduction refers to the degree to which an image conveys the luminance ranges of an original scene (or of an image to be reproduced in the case of digital imaging). According to Reilly and Frey, tone"is the single most important aspect of image quality." Tone reproduction is the matching, modifying, or enhancing of output tones relative to the tones of the original document. Because all of the varied components of an imaging system contribute to tone reproduction, it is often difficult to control.11
Resolution, dynamic range, real resolution, and tone reproduction combine to endow an image with quality. When defined and measured carefully, the terms can be used to describe the characteristics of an image, to compare quality characteristics of two or more collections of images, and to compare the digital image with its original source. The resolution and dynamic range values of a given image can also be combined to describe the size of an image in terms of the amount of data that is required to represent the image in digital form.
Describing digital objects. The description of an image or collection of images in terms of quality and quantity is but half the story of a digital image product. Equally important is the digital data that describe the digital object itself. In modern imaging systems, such descriptive data exist as a linkage of at least three components. The first are the technical data (often called the image header) that describe the format of the digital image and the ways in which the raw digital data are compressed to save storage space and transmission time.
The second component is data describing the characteristics of the digital object (which may consist of one or more digital images). Metadata is data about data and as such is fundamentally linked to the accessibility of an object. As mere bitmaps, digital images are stupid and cannot be found or understood without some level of metadata.
The third descriptive component is information that describes the relationships between or among digital objects. Structural indexes are a crucial component to any digital imaging system where the content is hierarchical in nature (such as archival collections, books, scrapbooks, classified photograph collections, and the like). It is a rare digital object whose accessibility cannot be enhanced through the use of structural indexes. Structural information may reside as separate data (e.g., an encoded finding aid) or be built into the metadata system itself (e.g., controlled subject headings in a bibliographic record).
In summary, at the heart of the digital world is communication, which cannot happen without a shared vocabulary and a shared system of symbols. Digital imaging is representation by numbers of the world we can sense (see, touch, hear, smell, and taste). Images as bitmaps are pictures without intelligence. All meaning embedded in the digital technology system derives from layer upon layer of numerical coding, most of which must be done by people rather than machines. In the end, then, digital imaging is more profoundly about us than about the tools we use.
We shall now turn our attention to the digital imaging processes and products by examining two general models.
Imaging Process Model
At its most elementary, the conversion of a book, a manuscript, a photographic negative, or a reel of microfilm is straightforward and linear. Source objects appropriate for conversion are selected and prepared for scanning; conversion occurs via scanning technology that transforms reflected light signals to digital data; access to the digital data is through display of the stored digital data. This apparent simplicity masks great complexity at all phases of the process.
Display technology is one of the main weak links in the entire system. Conversion technology is capable of generating far more data than can be usefully displayed by most of today's computer monitors.
Figure 7 is a schematic illustrating the elements of the process model. It is important to recognize that the complexity of a digital imaging system is only in part related to the complexity of the individual components. The elements of the process interact with each other to add complexity.
Imaging Product Model
The digital imaging process results in a product with its own characteristics that are distinct from the characteristics of the original sources. The biggest challenge in building an image product is to balance three issues: the characteristics of the source; the capabilities of the technologies of digital conversion; and the purposes or expected uses of the end product.
Figure 8 is a schematic defining the issues and suggesting a set of relationships that must be managed to produce an image product with sufficient value built in that it will be worth the cost and effort of ensuring its long-term preservation.
Of the three sets of issues (source, technology, uses) in the model, the concept of varying product uses is perhaps the least generally understood. A number of researchers in the field13 have begun to suggest that the quality of the end product can in some way be established through reference to one of three possible purposes that the product may serve for end users.
Each of these applications places separate, but increasingly rigorous demands on digital technologies. In each case, the use of an intermediate film or paper copy to facilitate the scanning process may be necessary or advisable. Finally, the disposition of original sources (including undertaking preservation treatments before or after conversion) is a separate matter. Ultimately, the purpose of the digital product is driven by access goals, while preservation of original source documents should be determined by the preservation needs of the original sources.
This leaflet has already suggested a number of issues with which librarians and archivists must wrestle if their digital initiatives will have lasting value. Here are five questions that transcend the specifics of digital imaging technology.
Written by Paul Conway