What is Kodak Photo CD Technology?
Description of the Study and the Methodology Used
Is Photo CD Technology the Right Scanning Method for You?
Assess Document Attributes
Control Photography and Digitization
Set Quality-Control Procedures
This brochure summarizes the findings of a study coordinated by Cornell University Library's Department of Preservation and Conservation that evaluated Kodak Photo CD technology as a tool for preserving and making available electronically a broad range of research materials. The project was funded by a grant from the New York State Education Department's Program for the Conservation and Preservation of Library Research Materials. It was carried out in cooperation with ten of the eleven comprehensive research libraries in New York State (the Big 11). (1)
Building on Cornell's decade of work on digital-capture requirements for text-based materials, the project evaluated Photo CD technology by controlling the factors that affect image quality during photography, digitization, and on-screen viewing. It sought to determine when Kodak Photo CD technology was the best solution for digital preservation and access, and what factors contributed to its effective application.
Our technical advisory committee included five nationally recognized experts in the fields of imaging science, micrographics, image quality, and standards development: Don Brown, imaging systems engineer, Eastman Kodak Research Laboratories; Michael Ester, president, Luna Imaging, Inc.; Steve Puglia, preservation and imaging specialist, National Archives; Jim Reilly, director, Image Permanence Institute; and Don Williams, imaging scientist, Eastman Kodak Research Laboratories. We are grateful to our skillful photographer John Pachai, of Cornell University Photography, who successfully implemented the photographic guidelines provided by Luna Imaging.
In addition to the Big 11 participants, Cornell recruited six other institutions to serve as beta testers of the applicability of our findings.(2) The assistance of both Big 11 participants and beta testers was vital in testing our methodology and benchmarking approach. Finally, we would like to acknowledge the contribution of the Eastman Kodak Research Laboratories, of Rochester, N.Y., in scanning the transparencies and creating prints from images.
This publication is directed especially to the staff of cultural, educational, and governmental organizations - such as museums, archives, and historical societies - that are considering using Kodak Photo CD technology to digitize their historically valuable materials. The brochure's three main goals are to:
Since 1992, Kodak's Photo CD technology has made it possible to digitize slides and transparencies, and to store images on Photo CD disks. Photo CD images can be viewed on television, using a Kodak Digital Science Photo CD player, or can be retrieved by most computers that are equipped with a CD-ROM drive. Photo CD technology is multifaceted, combining scanning technique, storage medium, image format and compression algorithm, and color management system for digitizing and presenting images.
Scanning and Storage
Transparencies and/or slides are scanned using a Kodak PCD Imaging Workstation (PIW), which consists of a film scanner, data manager, color monitor, CD writer, thermal printer, CD-ROM XA drive, and system software. Conversion is possible from either 35mm film, or from larger formats, such as 120mm or 4 x 5 sheet film. Although appropriate for 35mm, the Pro Photo CD is preferred for larger film formats, including 120mm and 4 x 5, as it provides a higher level of resolution. Photo CD disks are specifically designed for storing images from 35mm photography. (3) Some of the service providers offer both photo processing and digital conversion. And because service providers establish their own pricing schemes, costs vary from fifty cents to five dollars per image, reflecting a correspondingly broad range of image quality. Some offer value-added services, such as film cleaning, custom color-correction, encryption, and watermarking. Kodak Photo CD transfer sites can be located easily by using Kodak's Photo CD Transfer Sites search service. (4)
Image Format and Compression
Kodak Photo CD images are stored at five different resolutions, from Base/16 images at 128 x 192 pixels (for thumbnail views), to 16 Base full-resolution images at 2,048 x 3,072 pixels (Table 1). The Pro Photo CD disk adds a sixth level of resolution (64 Base) that provides high-resolution images (4,096 x 6,144 pixels).
TABLE 1: Resolution and Suitable Use of Kodak Photo CD Images
Photo CD images are stored in the ImagePac format (.pcd file extension), which is based on the Kodak Photo YCC color-encoding metric and Kodak PCD Data Manager. Depending upon the film type used and the detail and complexity of the image, for most 35mm images the size of an 16 Base ImagePac file is approximately 4.5 megabytes; a 64 Base file averages 21 megabytes. In the Photo CD compression scheme, image files are reduced in size to about a third of the original size without any visible loss of quality ("visually lossless"). The Photo CD ImagePac compression scheme removes some of the chrominance data, because the human eye is much more sensitive to brightness (luminance) than to color differences (chrominance). (5)
How to Find More Kodak Photo CD Information
Kodak Digital Science maintains an excellent Kodak Photo CD Products and Information Web site at <http://www.kodak.com/daiHome /products/photoCD.shtml>. It contains a general overview, sections on Web access, and downloading access software and technical papers. The Frequently Asked Questions link provides general information for those who are exploring the use of Kodak Photo CD technology. The site also has sections on transferring images, resolution, using and printing images, prepress, publishing, and professional photography. In addition, the Eastman Kodak Company maintains an on-line mailing list that functions as a forum for technical and nontechnical issues concerning Photo CD technology. (6)
Why Kodak Photo CD?
This study was inspired by the growing interest in the use of Kodak Photo CD technology within the library and archival community. The technology offers a commonly available and relatively inexpensive way to create digital surrogates of research materials. Many cultural and educational institutions have already invested in the creation of slide/transparency collections, and their use of these holdings as surrogates for digitization leverages on previous investments. To respond to questions surrounding the long-term accessibility of digital images, some institutions also view the slide/transparency intermediates as archival copies to be used as backups for the digital versions.
Each participating Big 11 research library contributed approximately ten documents from the past several centuries that fall into one of these graphic document categories: architectural drawings (e.g., blueprint, pen and ink), book illustrations, handwritten manuscripts, illuminated manuscripts, maps, works of art on paper (e.g., watercolor, sketch), posters, broadsides, playbills, scrapbook leaves, sheet music covers, and photographic prints. The documents represent typical graphic holdings that are considered essential elements of historical research for cultural institutions.
Cornell provided detailed guidelines to help the beta testers evaluate approximately ten documents and the corresponding images selected from their collections to be digitized via Kodak Photo CD technology (See Forms 4 and 5 on the Web site). The study involved only paper-based documents, and was limited to documents scanned using the basic Kodak Photo CD method (ImagePac). The beta-test verification process involved the examination of source documents and the on-screen displays of digital images (and, where possible, printouts of the images), and completion of the image-evaluation forms.
Because the original document is at the heart of Cornell's benchmarking approach, the evaluation process began there. Using forms developed by Cornell (Forms 1 and 4 on the Web site), the preservation administrators and curators of the participating institutions assessed the documents selected for the study to determine what information needed to be captured in digital images, such as type, medium, detail, color, highlights, shadows, depth, and color overlay. Other document attributes, such as document dimension, color information, and significant detail description and measurement were collected by Cornell staff members. These management data - along with the document assessments from the participating institutions - were recorded in a FileMaker Pro database.
The documents selected for the project by the Big 11 participants were sent to Cornell to be photographed by Cornell University Photography as 35mm color transparencies, using both negative and positive film. (7) The photographer followed guidelines from Luna Imaging, Inc. to produce optimum- and uniform-quality transparencies. (8) These guidelines recommended the film type to be used, lighting, alignment, placement of color targets, framing, placement of originals, and photographic processing.
Cornell was particularly interested in studying the effects of the photography and scanning processes on image quality. For comparison purposes, both the Eastman Kodak Company and Luna Imaging, Inc. digitized the transparencies, and adjusted color and density based on the accompanying color targets, the nature of the original documents, and film type. Images were recorded on master Photo CDs. Both negative and positive film were scanned to examine the difference in image quality between the two film types.
In addition, Luna digitized a representative group of twenty documents directly using a flatbed scanner (ScanView, ScanMate F8). They produced additional copies of the master disk, which were distributed to the participating Big 11 institutions for evaluation. These Photo CDs included the test set (scans from negative and positive film, and direct scanning) in addition to a number of resolution and color/gray-scale control images. Luna Imaging also included on the disk a copy of their Insight viewing software for convenient access to the images.
Cornell University Photography developed 8 x 10 prints from both negative and positive transparencies. Eastman Kodak produced 8 x 12 prints (on 9 1/2" x 14" paper) from the digital images, using a Kodak XLS 8600 PS dye-sublimation printer, controlled by color management software. Both the prints from the film and digital images were color-corrected based on the color and gray-scale targets included during photography.
Big 11 project participants attended a half-day workshop at Cornell that introduced them to the evaluation and benchmarking methods used in the study. Cornell prepared an image-assessment form (See Form 2) and sent it along with the original documents and the Photo CDs to the participants for the assessment process, which took place at each participant's home institution. The institutions were provided with detailed instructions on how to retrieve and evaluate the images, including recommendations for optimal viewing conditions (See Form 3). The preservation administrators from the participating research libraries examined and evaluated the original documents and the images for resolution, contrast, brightness, and color fidelity, and returned the completed forms to Cornell. Cornell's quality evaluation took place in the Preservation and Conservation Department's Digital Imaging Unit and was based on twenty representative documents. Form 6 contains a list of hardware and software used during the evaluation process.
Kodak Photo CD technology is a viable means to digitize many, but not all, categories of special collections materials for access purposes.
In addition to resolution and color assessment, the project staff, Big 11 participants, and beta testers evaluated sixty-seven Kodak Photo CD images in terms of their overall quality, and rated them in one of four categories:
The participants ranked none of the digital images as providing improved quality over the original documents, but they did find that in 85 percent of the cases, the Kodak Photo CD images represented acceptable digital surrogates for the originals. Table 2 shows the percentage distribution of overall image quality across document categories, while Table 3 provides the numeric distribution by document type. These evaluations are highly subjective, and the figures were considered in tandem with evaluator comments in reaching our conclusions. For instance, several reviewers considered some images adequate or even comparable to the originals even though they were not able to read the finest text contained in them, while other reviewers rated similar images as unacceptable. Nonetheless, some general conclusions can be derived based on these rankings:
TABLE 2: Overall Evaluation of the Images Included in the Study
TABLE 3: Evaluation of the Images by Document Description
Although this study did not include a user-needs assessment, the following questions can be posed and examined in an attempt to evaluate Photo CD as a potential tool to digitize your collections:
Cornell has spent several years developing ways to correlate document attributes with digital outcomes, called benchmarking. Benchmarking is a systematic way to forecast a likely outcome of using digital imaging for preservation and access purposes. The process begins with the assessment of documents, considers variables associated with digitization, involves the use of formulas, and requires a means for confirming results (Kenney and Chapman 1995, 1996). Benchmarking enables institutions to make informed decisions prior to committing their resources to a digital conversion project. The approach can be used effectively with the Kodak Photo CD technology to predict whether satisfactory results will be achieved when digitizing a document.
The first step in benchmarking is to characterize your document's attributes. The document-assessment forms used during the study (See Forms 1 and 4 on the Web site) will help you assemble the document data that are essential for determining an appropriate scanning technique for your collection. Because it may not be practical to fill out a form for each document (especially if you have a large collection), it is important to choose representative documents that are typical of your collection, and collect profile data for each unique type.
Resolution is one indicator of image quality. To determine whether the effective resolution achieved via Kodak Photo CD will meet your needs, consider the following factors:
These six factors will affect the representation of detail in the digital file. Let's see how they are interrelated.
Pixel Dimensions of the Kodak Photo CD
The Kodak Photo CD process can be thought of as superimposing a grid of 2,048 x 3,072 pixels (dots) on the document. This grid is used to produce a sampled version of the original. For small documents, this grid is "compressed" to cover a small surface area: pixels are spaced closer together, effectively increasing the resolution (dots per inch [dpi]). Small documents with fine details can be satisfactorily rendered using Kodak Photo CD technology. With larger documents, each dot represents more area, which reduces the resolution and consequently the ability to record fine detail. For instance, a 4" x 6" document will have an effective dpi that is ten times as great as a 40" x 60" document, when captured with the Kodak Photo CD scanning array.
Remember that with Kodak Photo CD, a film intermediate is scanned instead of the original. On film, the two documents appear to be the same size - the only difference is that the 40" x 60" document will have to be reduced ten times more than the 4" x 6" document to fit on the film. The degree of miniaturization (known as the reduction ratio) has the same impact on the recording of fine detail as does the difference in physical dimensions. As the reduction ratio increases, details in the original source are made smaller; consequently fewer pixels are used to record detail. In effect, the resolution (or dpi) goes down.
If a source document is large and it has large features, then the lower dpi is not a problem, because the larger details can be resolved at lower resolutions. This is often the case with posters, which have large features because they are meant to be viewed from a distance. The problem comes when the size of the document is large in relation to the level of detail presented on it - the most obvious example is an oversize map that contains fine lettering. In determining whether Kodak Photo CD is a viable scanning option, one must take into consideration the limitations of the grid of dots that the scanning array represents; compare it to the dimensions and level of detail that comprise the source document; and ensure that the document is properly aligned during filming.
Document Alignment for Photographing
When documents are digitized via Kodak Photo CD, a 35mm film intermediate is created, which is then scanned. To ensure that the maximum resolution is achieved in this process, one must consider how the document is placed or aligned for filming.
To determine the best alignment for filming - ensuring that the entire document is filmed and that as much of the frame is filled as possible - one should relate the aspect ratio of the film to the aspect ratio of the document.
* The aspect ratio of the 35mm film frame.
A document, a film transparency, and the scanning array used in the Kodak Photo CD process are rectangles, which can be characterized by their aspect ratio. Aspect ratio is a way to relate one dimension to another, and is used to define the shape of a rectangle. In the case of a 35mm film transparency, we know the dimensions - on the short side the film measures 23.33 mm; on the long side 35 mm. The aspect ratio is 35/23.33, or 1.5. The resulting rectangle of the film transparency is one-and-a-half times longer than wide. The Kodak Photo CD process uses a scanning array of 2,048 x 3,072 pixels to create a digital grid, or rectangle, with the same aspect ratio as the film transparency. Because the aspect ratio of the film transparency and the Kodak Photo CD scanning array are identical, the film can be aligned perfectly for digitizing.
In a document, the aspect ratio relates the height to the width. Because an image can be rotated after scanning, the orientation of the document for filming can be either portrait or landscape. The document's aspect ratio is determined in the same way as for the film, by taking the larger dimension and dividing it by the smaller one.
Ratio of Detail Size to Original Document Dimension
The original size of a document affects the capture of fine detail. We can characterize the magnitude of that effect by calculating a ratio of detail size to physical dimension.
The first step is to locate and measure a detail - the smallest complete part of a document that is considered essential to the document's meaning. For textual materials, this is usually the smallest lowercase "e". For nontextual materials, it might be a person's eye in a portrait, a single pearl in a woman's necklace, or a small bird in the background of a painting. In measuring detail, it is best to find a very sharp or crisp detail rather than one that is fuzzy or poorly defined. To measure the significant detail, use a loupe, or a microscope that can resolve details as small as 0.1 mm. Next, measure the document's height and width, then determine its aspect ratio and the preferred alignment for photographing, following the guidelines above. Form 4 on the Web site provides instructions for measuring details and document dimensions.
Assuming that you will optimize the placement of the document for photography, you can then determine the detail-to-dimension ratio by using the following formula:
ratio of detail to document size = (x x 0.039)/y,
This ratio can also be characterized as a percentage by multiplying the results by 100. In general, if the percentage is greater than 0.78 percent, then the effective resolution achieved via Kodak PCD should prove excellent in rendering detail in the source document. This prediction is based on the assumption that the document is filmed and scanned properly.
We predict that if this document is digitized with Kodak Photo CD, the detail will be rendered with excellent results. [Sixteen pixels would span the height of the "e".]
If detail is difficult to characterize, the thinnest stroke can also be measured and used to benchmark whether Kodak Photo CD will be an appropriate conversion process. A stroke width is the thickness of the finest significant line in a document, such as the thinnest line used to render a handwritten letter or the thickness of a telephone wire in a photograph. The stroke width can be used for "x" in the formula above. Kodak Photo CD should provide excellent rendering for any document whose stroke-to-dimension ratio is greater than 0.1 percent. Again, this prediction is based on the assumption that the document is filmed and scanned properly.
We would predict excellent rendering of that stroke, because more than two pixels would span the stroke width.
Levels of Resolution Quality: Dots per Detail and Dots per Stroke Width
The formula above provides an easy, generalized way to predict whether Kodak Photo CD technology can result in excellent rendering of detail. But what about those cases where the percentage ratio is at or below the recommended level? Consider, for instance, if your document's width were 20" instead of 10". The detail-to-dimension ratio would be 0.39 percent, less than the recommended 0.78 percent. You may still produce satisfactory results using the Kodak Photo CD, but must now make an evaluative judgment of what constitutes acceptable quality. Cornell has devised the formulas and quality-projection tables (Tables 4 and 5) to assist you in making these judgments.
The first step is to determine the number of dots that will be used to render the height of a detail. We will begin with the ratio formula, (x x 0.039)/y, which relates detail to the document dimension to be aligned with the film frame. We can extend that formula to include the scanning process itself, and derive an estimate for the number of pixels that will be used to record detail.
We know that Kodak Photo CD uses an array of 2,048 x 3,072 pixels to scan the film. Because we determined which dimension of the document to align for filming, we also know the corresponding pixel dimension that will be used to record that dimension. If we align the document by its smaller dimension for filming, we know that up to 2,048 pixels will be used to record that dimension in its miniaturized version on the film. If we align the document by its larger dimension for filming, we know up to 3,072 pixels will be used to record that dimension from the film. In our last example, we aligned 20" with the 2,048-pixel dimension, achieving a resolution of 102.4 dpi (2,048 divided by 20).
With this information, we can calculate how many dots will be used to record detail by multiplying the detail-to-dimension ratio by that pixel dimension.
Let's return to our example. Using this formula, we can calculate that approximately eight pixels will be used to render the height of the smallest detail for a document whose smallest dimension is 20": ((2 x 0.039)/20) x 2,048 = 7.98.
dots per detail = ((x x 0.039) /y) x z,
Will this be enough to render the detail in a satisfactory manner? Figure 2 illustrates the effects on character representation based on the number of pixels used to render their height via Kodak Photo CD. We have concluded that a minimum of eight pixels are necessary to represent detail in a consistently acceptable or just-legible manner. Twice that many pixels should provide excellent fidelity to the original, taking into consideration color effects, document density and contrast, original detail rendering, and the quality and condition of the film intermediate.
The formula can also be used to calculate the number of pixels that will be used to render the finest strokes, when "x" represents stroke width. Figure 3 shows that approximately two pixels have adequately rendered the stroke. Whereas, in Figure 4 the borders of the document were inadequately rendered as there was at most a single pixel straddling the fine line (stroke).
Our assessment of these characters and other details was based on reviewing documents and digital images included in the Kodak Photo CD project, and largely confirmed by others at the participating and beta-testing institutions. The chart presented below is not definitive but rather is an attempt to provide a general means for equating pixels to perceived levels of quality, and builds on work conducted earlier at Cornell and elsewhere. (9) The quality rankings are conservative because they are meant to be applicable to a range of document types and to factor in the effects of color, density, contrast, and generational loss. It may be possible, for instance, that excellent detail rendering may be achieved on some documents at lower pixel counts, or that a stroke can be rendered acceptably at lower levels (for instance, at one pixel), but that details may not be (for instance, at seven pixels). Calculating both stroke and detail pixel counts, and comparing the results, may be the best way to predict whether Kodak Photo CD will provide sufficient resolution. These recommendations are for benchmarking purposes only. Because image quality is affected by several variables, these benchmarks should be confirmed by evaluating the output of representative samples of the material to be scanned. (10)
Table 4 shows a quality index that may be used to forecast levels of detail quality which can be achieved via Kodak Photo CD (see Figure 2).
Table 5 shows a quality index that may be used to forecast levels of stroke quality which can be achieved via Kodak Photo CD.
TABLE 4: Quality Index for Detail Rendering
questionable, should be assessed on screen
TABLE 5: Quality Index for Stroke Rendering
questionable, should be assessed on screen
Generational Loss Caused by the Number of Links in the Digitization Process
Each component in the digitization process will affect in some way the resulting quality of the digital image. In the Kodak Photo CD process, the digital image is two "generations" away from the original: it has been copied to film (first generation), and that film has been scanned (second generation). At each step in this process, there is a loss of quality, which can be seen in both the representation of color appearance and the rendering of fine detail (Figures 5 and 7). It is possible to mathematically determine the effects of the total imaging chain on the final resolution (ANSI/AIIM 1993), but our formulas are based on perceived quality of the digital file and thus have already factored in those effects. We will discuss issues associated with color appearance later.
Two observations bear on this problem. First, the resolution of the Ektachrome positive film used in this project was fairly low, averaging 55 line pairs per mm. The impact on image quality is most evident in filming large documents that include fine detail - as the reduction ratio increases, the film can record less fine detail. Generally, the loss will be tolerable if documents with details averaging 1 mm in size are reduced 10 times or less. Because there is an additional drop in quality when going from film to digital, however, one must consider how much quality is compromised in the filming process, and consider using larger-format films (e.g., 70mm, 105mm, or 4 x 5 transparencies) and the Pro Photo CD process for oversized documents that contain fine detail.
Second, there is a discernible drop in resolution from the 35mm film to the digital image created using the Kodak Photo CD process. We observed a film-to-digital drop-off that ranged from 5 to 20 percent, with the average being approximately 14 percent. We used the RIT Alphanumeric Resolution Test Object to determine how much resolution was lost in the transition from film to digital. (11) Figure 6 compares RIT rendering from two vendors, with two film types and a direct scan.
A range of factors can affect how well the RIT target can represent this loss, and in the future we will rely on the use of specially designed technical test charts and software programs to compute the Modulation Transfer Function. Our findings, therefore, are not precise, but do represent a general trend. Resolution loss may be evident on-screen, but may not be discernible in printouts made from the digital files.
Nature, Quality, and Condition of the Original and the Film Intermediate
The perceived quality of detail rendering can be affected by such factors as the nature and complexity of the document, detail and background color, document density, color contrast, and subtleties of shading. Evaluation can be very subjective, differing from one viewer to the next. The shape and characteristics of a detail will also affect how well one judges its digital representation. It is often difficult to measure and assess detail in photographs because they do not exhibit the same sharpness or crispness as graphic or textual materials. Details in the highlights and shadows are more difficult to judge than those in the middle tones. These factors are reflected in project evaluations: for example, reviewers were more critical in evaluating resolution of manuscripts than of photographs, where detail rendering is often difficult to assess (see Table 3). Similarly, the film intermediate can introduce distortion if the lens is not properly focused, the lighting is poor, or machine vibration occurs at the time of image capture. Although these differences were noted, we did not come to any general conclusions about how such factors affect detail perception, only that they did. This is why the formulas and quality indexes given above err on the side of the conservative, and have been developed with a range of document types and conditions in mind. More research needs to be conducted in this area.
From Source Document to Photographic Intermediate
Because the Photo CD technology does not involve direct scanning of source material but rather relies on the creation of 35mm film intermediates, the quality of both photography and film processing can greatly affect the quality of digital images. When attempting to determine how well the Kodak Photo CD process can render information present in source documents, controlling the variables associated with photography is an important consideration. As Michael Ester states (1995, p. 148), "if visual detail subtlety is not in the photographic medium, neither will it appear in the digital image." We offer the following recommendations to maximize photographic quality:
Good Photography Great care should be taken in the creation of the transparencies to minimize image degradation. The photography guidelines provided by Luna reiterated elements of good studio practice.(12) Therefore, photographers who are embarking upon Photo CD projects do not need to acquire a brand-new set of skills, but rather should build upon the traditional "good photography" practice.
Our photographer's "potential pitfalls" list is based on his Kodak Photo CD project experience:
As elaborated on in the "Use of Targets" section below, the photographer must use a gray scale during photography, to be able to compare the tone values of the original and the transparency, and to set the exposure and film processing steps in the photographic environment accordingly.
Resolution: Maximize Digital Real Estate! We have seen that resolution is one of the indicators of image quality. As a general rule, the higher the resolution of an image, the greater its clarity and definition. Effective resolution achieved via Kodak Photo CD depends on original document dimensions: as the size of the original increases, effective resolution decreases. When capturing a document, it is important to minimize unnecessary background and fill the frame as much as possible with the document to utilize the full scanning array. Michael Ester, who coined the term, "digital real estate," recommends cropping out extraneous content in the photographic step prior to scanning.
Film Stock Both Luna Imaging and Eastman Kodak discourage the use of duplicate film, to avoid the quality degradation resulting from second-generation film (Ester 1995). High-quality scans are possible from either negative or positive transparencies. Kodak and Luna recommend the use of negative film, because it has a smoother response curve over its dynamic range and it produces greater shadow and highlight detail.(13) This recommendation is also substantiated in our study. The digital images created from negative film exhibited marginally better resolution but greater contrast (see Figure 7). Fine-grained, slower negative film is preferred, as coarser film grain can be very noticeable when images are examined at full resolution. Grainier, high-speed films do not compress as efficiently as slower-speed films, because the grain may even become a detail in the images, occupying pixels.
Film Processing Not only photography but also film processing can have a great impact on the quality of transparencies and slides. Using a service provider that offers both film processing and image-transfer services may be less expensive, but it is important to ensure that both processing and scanning processes will be performed by qualified, trained operators.
Ester (1995, p. 150) warns against "darkroom magic," arguing that correcting images based on color bars during processing might cause certain parts of the image to look wrong. The corrections should be implemented either during the photographic process (via rephotography), or after scanning, using image-enhancement software. Existing image enhancement tools allow easy and quick image modifications, such as color correction and sharpening. Master copies of images should be saved to maintain a copy of the unaltered image for future applications.
Ester (1996, p. 13) recommends "matching to the scene" rather than "matching to the film," to ensure that the image will be closer in appearance to the original document, not to its film surrogate. He also cautions that not every image can be processed to match the scene. For example, correcting some fine-line drawings that were photographed with slightly overexposed areas might wash out detail.
It is easier for the service provider to transfer the images to the Photo CD right after film processing, before the film has been cut into shorter strips or the slide film is placed into individual slide mounts, which cover approximately 5 percent of the film area. If scanned with the slide mount in place, the scanned image may show the black outline and rounded corners of the mount frame.
Test Before You Proceed Preliminary production tests are highly recommended to catch problems in a timely fashion. During our project, after the first twenty documents were photographed and processed, the color transparencies and the originals were sent to Luna Imaging for scanning. Upon inspection and comparison of images and originals, a red cast was found that occurred during photography (see Figure 8). Upon diagnosing the problem, our photographer experimented with different red filters and managed to eliminate the cast.
Documentation We recommend that you document the equipment and settings used during photography: camera, lens, lamps, filters, light meter, densitometer, control bars used, film type, and processor used. Even if this information is not necessary during your current project, you may need the specifics of photography in the future for other purposes. If more than one transparency is created for each document (e.g., shooting the same document using different settings, lighting, and polarization for optimal capture; or placing them on dark or light surfaces), clearly identify the film selected for scanning, and ideally separate the film that will not be scanned from the batch being sent to the service provider.
Use of Targets
Cornell used several color and resolution targets to serve as checks on the photography, to provide measurements for the transparency scanning, to calibrate monitors and printers, and to facilitate visual evaluation of on-screen and printed versions of the digital images.
Color and Gray-Scale Targets It is highly desirable to include gray-scale and color targets during photography to provide a level of control over color and tone shifts. (14) The color-control bars help the photographer and scanner operator to compare the color of the subject with known printing colors. The gray scale is a quality-control device of stepped, neutral values. Both color and gray-scale targets are used in color-matching images during photography, digitization, on-screen viewing, and printing.
Good color-matching requires not only the information on the color scale, but also the gray-scale control of shadows, midtones, and highlights. Don Brown from Kodak, one of the members of our Technical Advisory Committee, says "if you get your neutrals right, most of your color management work is done." We recommend that even for color documents, you use both color and gray-scale targets. Ideally, these targets should be photographed with the individual document. Alternatively, provided that the same lighting and focus are used throughout filming and the film is processed in the same batch, these targets can appear at the beginning of each film roll. Photographers should be provided with detailed instructions on how to place targets to maintain consistency, and to minimize wasted pixels.
During our study, the color targets were photographed with the individual documents, frequently overlaid to cover certain sections of the documents. Because of the research nature of the study, no attempt was made to strategically place the control bars to avoid their overlap with the original documents. In a production setting, the placement of the targets will depend on the aspect ratio of the document. If the aspect ratio of the original document and Kodak Photo CD (aspect ratio of 1.5) are not the same, the control targets could be placed over the extraneous space left during photography, as shown in Figure 9. However, if the aspect ratios are equal (or very similar), there may not be any extra space for the control bars. In these cases, the smallest available bar should be used, and even including a thin strip of the target will be sufficient.
As illustrated by Columbia University's experience during Phase II of the Oversize Color Images Project (Cartolano, Gertz, and Klimley, 1996), if an automated script will be run to crop the control bars after digital images are created, the control bars need to be placed consistently to facilitate the automated cropping process. The color-control bars should be kept intact in the master file for future use.
Color and gray-scale control bars will fade over time, and need to be replaced periodically. They need to be stored away from direct light sources, in folders or envelopes. If the targets are used often, they need to be replaced with new ones on a regular schedule, such as once a week.(15)
Resolution Targets Resolution targets can be used to measure and evaluate digital resolution, provided they have been photographed and scanned at the same settings as the originals. (16) However, most resolution targets were developed for the micrographic or photographic industries, and will not consistently predict how well a digital system has resolved detail at various spatial frequencies. These targets are useful only for characterizing where the system fails, but even here the results may be inconclusive, due to variations in lighting, target density, and the introduction of sampling errors at the point of digitization. The most effective target to use for calculating the loss of resolution from the original is the RIT Alphanumeric Resolution Test Object.
A more reliable way to define resolution loss is the Modulation Transfer Function (MTF), a mathematical method for characterizing how well detail is preserved across a broad spectrum of frequencies. MTF has been used extensively in industry, but the means for employing it outside special research labs has not been available until very recently, and therefore it was not practical to utilize in our study. As elaborated by Williams (1998), the technique is reaching a stage that will allow its general use in the very near future. The Library of Congress, for example, has begun to define MTF requirements for the digitization of pictorial materials (Library of Congress 1997).
The equipment used, its performance, and operator judgment, can have a tremendous impact on determining image quality. In our study, each of three scanning technicians produced slightly different results. One made an extra effort to match colors based on the control bars present in the transparency, in addition to using sharpening tools during scanning. In addition, Luna Imaging included a scanned Kodak Q-60 Color Input Target with every disk as a standing check on input calibration, and as a reference for subsequent changes to the images (including transfer to a new computer platform).
Although the hallmark of the Photo CD system is the control of color and exposure to produce optimum scanned images, a skilled scanning operator can improve the results. Data-manager and color-monitor components of the Kodak PCD Imaging Workstation (PIW) help the scanning technician make adjustments to the image. Additionally, how an individual service provider maintains the equipment and handles the original film can vary greatly, affecting the final quality of images. For example, lack of color calibration of scanning devices will have a major impact on the scanned images. Don Brown, of Eastman Kodak Research Laboratories, recommends that all Photo CD service providers adhere to the following basic scanning practices:
Brown also suggests doing a first-approximation calibration to improve the quality and consistency, one that is based on the appearance of the Kodak Gray Scale captured during photography. This three-step technique can be summarized as follows:
Form 7 provides the details of this technique.
This technique assumes that the gray scale is perfectly neutral, the copy stand lighting is fully uniform, and the frame-to-frame exposure variability of the film is negligible. All of these are reasonable assumptions if care was taken during the photography, because the scan-to-scan variability of the scanner is known to be very small. Individual scans can be custom-adjusted with this same technique, but the goal is to create a "standard correction profile" for all images at the beginning of the scanning session, so that they can be automatically applied to all subsequent images. The PIW software accommodates such a work flow, and after the initial setup, the scanning process can be very efficient. When following this technique, gray-scale patches on a few subsequent images should be examined to ensure that the image on which the corrections were based is representative of the entire series. Once this confidence is achieved, the rest of the scanning can be done with no operator adjustments. If anything changes within a series of images (such as photographic lighting difference, film emulsion change, even film processed at a different time), it is a good idea to repeat the correction for each subseries within the entire series of images.(17)
Based on our experience, we offer the following additional tips:
The impact of the image-retrieval process on perceived image quality is often underestimated. If retrieved in improper image-display conditions, even a very high-quality image may come across as unsatisfactory. For example, viewed using an underpowered computer that can not provide a full palette of colors, a 24-bit color image might look heavily "posterized," losing the beauty and variety of the colors present in the master image.
The factors that affect the perceived image quality on-screen include:
We recommend the following hardware configuration to optimize the viewing of Photo CD images:
Image Retrieval Software
Several types of freeware- and shareware-viewing software products are available on the Web. To optimize images, we suggest that you use Adobe Photoshop with the Kodak Photo CD Acquire Module plug-in to ensure correct mapping of Photo CD colors. The Kodak Photo CD Acquire Module plug-in allows you to select, customize, and import Photo CD images into Photoshop, utilizing the KODAK Precision Color Management System (KPCMS). Directions on how to download and install this free plug-in are provided in Form 3 [Appendix C (Opening and Printing Images Using Kodak Photo CD Acquire for Adobe Photoshop)].
Eastman Kodak has recently announced the availability of a Photo CD Java Applet for Java browsers (e.g., Netscape Navigator 3.0) to enable the retrieval of native ImagePac images on the Web.(20) Before the availability of this Java Applet, ImagePac images needed to be converted to one of the commonly supported Web image formats such as GIF or JPEG, because the ImagePac format was not supported by any of the commonly used browsers. This new applet allows the images to be viewed using functions such as <zoom in> and <zoom out>, and <pan> for higher-resolution images. Another advantage of this Web format is its color management component. Unlike GIF and JPEG files, Photo CD images are stored in Photo YCC color space, which is based on international color standards and therefore can comply with the International Color Consortium's (ICC) color profile (for further information on ICC, see the Color Management Systems section). Developed collaboratively by Kodak, Hewlett-Packard Company, Live Picture, Inc., and Microsoft Corp., the FlashPix file format is emerging as an alternative to ImagePac, because it is designed specifically for Internet applications.(21) Images in FlashPix format files are stored at multiple independent resolutions, and each resolution is made up of discrete square tiles. These features allow users to choose an appropriate resolution consistent with their needs, and to access directly (and quickly) the specific areas of an image.
An important prerequisite for viewing images is a controlled viewing environment. The monitor and the original document require distinct viewing environments. The original is best viewed in a bright surrounding, and the monitor works best in a low-light environment. However, a "low-light environment" does not indicate a dark room. Viewed in the dark, the image on the monitor would appear to be deficient in contrast.
To create an ideal setup:
The monitor should be brighter than any other light source in the room. With this in mind, you may want to separate the monitor and the viewing space for examination of the original. You should give your eyes some time to adjust when switching between the two environments.
If you are comparing originals to digital images, a color-viewing booth, ideally one such as The Judge II from GretagMacbeth,(22) should be used for visual evaluation and comparison. If you are using a viewing booth, the document should be tilted, to eliminate reflections and glare from the lights in the booth. Set the lighting of the viewing booth at Noon Sky Daylight at 5,000 Kelvin (the ANSI standard). If you do not have a viewing booth, place the original document in a position that minimizes reflections and glare, preferably under a lamp that uses a natural-daylight fluorescent bulb, which will allow you to control the lighting of your viewing environment.(23) The colors of your clothes may reflect on your monitor and can affect your color perception. Therefore, we suggest that when working, you wear neutral colors, such as gray, black, or white.
Images may appear to be different when viewed on different monitors. Calibration is the process of adjusting your monitor's color-conversion settings to a standard, so that the images displayed on a variety of monitors look the same.
The ideal calibration method uses monitor-calibration hardware and accompanying software. However, if you do not have access to the necessary hardware and software, you may want to use your application program's calibration tools. For example, Adobe Photoshop includes a basic monitor-calibration tool, which can be used to eliminate color cast in your monitor display, ensure that your monitor grays are as neutral as possible, and standardize the display of images.
The keys to calibrating a monitor are to set the gamma and white point. You can optimize your monitor performance by using the following suggested settings:
gamma = 2.2
white point = cool white (5000 Kelvin)
color depth = millions of colors (24-bit)
Kodak provides a reference image to evaluate the color calibration of the monitor and/or image-viewing software used. It can be downloaded free from:
<ftp://ftp.kodak.com/pub/photo-cd/general/referenceImage.sea.hqx> (for Macs)
<ftp://ftp.kodak.com/pub/photo-cd/general/referenceImage.exe> (for PCs)
The gray-scale reference image shown in Figure 10 was created through purely digital means and did not involve scanning film on a Photo CD imaging workstation. The patches represent ideal neutral patches in Photo YCC color space. This reference image can be used to evaluate grays in your image, to assess how well grays have been replicated in your image. It can also be used as a tool to adjust colors of the image (not the monitor). Form 3 [Appendix F (On-screen Color Quality Assessment)] describes how this reference image can be used to evaluate grays in your image.
In addition to calibrating your monitor, follow these simple recommendations to optimize your monitor setup:
Color Management Systems (CMS)
One of the main challenges in digitizing color documents is to maintain color appearance and consistency across the digitization chain, including scanning, image display, and printing. The principal source of difficulty in accurately reproducing colors stems from the fact that the color gamut of input and output devices will differ because of a lack of common color space. For example, scanners capture colors using the RGB (red, green, and blue) or Photo YCC color model, while printers use the CMYK (cyan, magenta, yellow, and black) model that is based on the absorption of ink into paper.
Examples of CMS software include Colorific, Apple ColorSync, and Kodak Precision Color Management System (KPCMS). Adobe Photoshop (Versions 2.5.1 and 3.0) includes KPCMS as an extension that accurately reproduces Kodak Photo CD color space. In addition, the Acquire Module plug-in takes advantage of the KPCMS, and provides a color managed conversion to various output devices. The December 15, 1997 issue of RLG DigiNews provides a technical review of CMS, including a list of most popular stand-alone CMS software products (Technical Review 1997).
The International Color Consortium (ICC), established in 1993 by eight industry vendors, is intended to create, promote, and encourage the standardization and evolution of an open, vendor-neutral, cross-platform color management system architecture and components. Their "International Color Consortium (ICC) Profile Format" is intended to represent color consistently across devices and platforms. According to Reils (1997), ICC specifications may be slow in coming but they are already revolutionizing CMS, because they provide basic guidelines describing what should be included within input and output device color profiles, to ensure proper mapping to an ICC-compliant utility or application.
In addition to CMS, which provides an integrated tool, there are a wide range of independent color quality-control instruments, such as densitometers, colorimeters, and spectrophotometers. These devices provide full spectrum measurement and matching of colors from a wide range of objects, reflective surfaces, and monitors.
Image Quality Assessment
Once your viewing environment has been controlled, on-screen image quality evaluation can proceed. You will want to evaluate images in several ways: overall presentation, detail capture, color appearance. Image-quality assessment can be a subjective process, and we recommend that all images be evaluated by the same person, using the same equipment and settings, to minimize factors introduced by these variables. The Image Assessment Survey used in this study and the accompanying appendices, available in the electronic version of this brochure (Form 2 and Form 3), can serve as guides to inspecting and evaluating detail, color appearance, and overall presentation.
Legibility and Detail Examine the fine details in your document and its image counterpart to check if they are properly rendered in the light and dark portions of the image. Count the pixels that comprise detail and stroke by zooming in until you can identify pixels as individual rectangles. After counting, evaluate whether the detail or stroke has been adequately represented (see Form 3 [Appendix E: Inspection of Legibility and Fine Detail]).
Color Compare the digital image and the original document to see that the colors and tones of the original document have been successfully reproduced. Note if there is an overall color shift, which can easily be modified, or a range of color shifts that will require image-by-image adjustment.
Overall Evaluation To assess the quality of your image as a whole, ask yourself the following questions:
Printing was not a significant part of our study. However, based on our limited experience with printing, we offer these recommendations:
This demonstration project led us to as many questions as answers. Additional research on the effectiveness of Kodak Photo CD in digitizing other source materials, such as three-dimensional museum objects or paintings that exhibit overlay, depth, and translucency characteristics, is advisable. A logical follow-on study might be to compare the quality, cost, and utility of creating Kodak images from a range of sizes of photo-intermediates, including 4 x 5 and 120mm transparencies. The current project barely touched on issues associated with printing or Web access, and did not address database, indexing, and other organizational considerations.
Nonetheless, this project has helped Cornell evaluate the use of Kodak Photo CD technology for digitizing special collections materials. Our main conclusions - involving digital benchmarking, ratio of physical size to detail, controlling the digitization chain, optimizing image acquisition and display, and establishing a quality-control program - have been summarized in this brochure, and offer a good starting place for those contemplating a Kodak Photo CD project. We invite our readers to continue their education by linking to the numerous Web sites we have listed.
1. Columbia University, Cornell University, New York Public Library, New York State Library, New York University, State University of New York at Binghamton, State University of New York at Buffalo, State University of New York at Stony Brook, Syracuse University, University of Rochester.
2. The beta testers for the Kodak Photo CD project included staff members of Carnegie Mellon University's Hunt Institute for Botanical Documentation, Harvard University Library's Map Collection, the Northwestern University Library, Princeton University's Department of Art and Archaeology, the State University of New York at Stony Brook's Map Collection, and the University of Michigan's Museum of Art.
3. Kodak Photo CD disks were developed by Eastman Kodak Company as a medium to store and play both multimedia presentations and digital photographs.
4. Visit their Web site at <http://www.kodak.com:80/digitalImages /piwSites/piwSites.shtml>, or telephone 800 939-1302.
5. All color scanners, including Photo CD scanners, digitize images with red, green, and blue filters (based on the RGB color definition). Once the scan is made, Kodak PCD Data Manager transforms the RGB data into three color components named "Y", "C", and "C": one channel of luminance (Y) and two channels of chrominance (C). Because Photo YCC color space is based on the broadcast-television industry's YIQ technique, Photo CD images can be viewed both on television and computer.
Technical papers on Kodak YCC can be found by clicking on "Photo CD Technical Papers" on Kodak's Photo CD Products and Information Web page: <http://www.kodak.com/daiHome/products/photoCD.shtml>.
For a comprehensive discussion of color models and other issues surrounding color appearance, see GretagMacbeth (1997).
6. To join this mailing list:
1. Send e-mail to: firstname.lastname@example.org.
2. Do not include a signature or subject line.
3. In the body of the message, type: "subscribe photo-cd <First name> <Last name>" (For example: "subscribe photo-cd John Doe")
4. When posting a message to the list, send e-mail to: email@example.com.
7. The original documents were photographed using a Marron Carrel Model 14000 camera, Kodak Royal Gold 25 (formerly Ektar 25) and Kodak EPY (64T) films, a Micro-Nikkor lens, and Q750CL/DC Quartz lamps.
8. Luna Imaging, Inc. (1315 Innes Place, Venice, CA 90291, 310 452-8370) offers a range of products and services designed to meet the needs of the cultural-heritage community, including digital conversion and processing, and image-server software that can be integrated with existing database systems.
9. See for instance, Kenney and Chapman (1995). One can apply this approach to the recommendations for color scanning from other institutions. Cartolano, Gertz, and Klimley (1996) suggest that a 1 mm-high character can be rendered with excellent quality at 200 dpi in 24-bit color. At 200 dpi, a 1 mm-high character would be represented by 7.8 dots. Cornell's findings indicate that this would provide satisfactory, but not fully faithful rendering, of detail. Both the National Archives and the Library of Congress recommend scanning maps at 300 dpi to meet access purposes, which would spread approximately twelve dots across a 1 mm-high character.
10. It should be noted that on-screen display of detail and stroke may be misleading, because of color "bleed" or spread to surrounding pixels. Reviewers consistently over-estimated the number of pixels observed to cover strokes and details on screen.
11. The 3 x 3 version, 1.9 density, A200-610-D1.9, is available from Rochester Institute of Technology, T & E Center, at 716 475-2739.
12. The Luna Imaging photography guidelines included recommendations related to film type used, lighting, alignment, placement of color targets, framing, placement of originals, and photographic processing. These general guidelines are also presented in Ester (1995).
13. Click on "Transferring Images" at Kodak's Digital Science Photo CD Technology FAQs Web page, <http://www.kodak.com/global/en/service/faqs/faq1001a.shtml>.
14. The following targets were used during the study: Kodak Q13 and Q14 Color Separation Guide and Gray Scale, Kodak Q-60 Color Input Target, Kodak Neutral Card, RIT Alphanumeric Resolution Test Object, AIIM Scanner Test Chart, IEEE Std 167A.1-1985. Ordering information is provided in Form 6.
15. From the user guide packaged with the Kodak Color Separation Guide and Gray Scale Q-13.
16. The following targets were used during the study: Kodak Q13 Color Separation Guide and Gray Scale, Kodak Q-60 Color Input Target, Kodak Neutral Card, RIT Alphanumeric Resolution Test Object, AIIM Scanner Test Chart, IEEE Std 167A.1-1985.
17. Kodak's Photo CD Technical Papers Web page <http://www.kodak.com/productInfo/technicalInfo/photoCDPapers.shtml> includes several documents that can help you optimize the performance of Photo CD Imaging Workstations (PIWs). Of particular utility is Fully Utilizing Photo CD Images: Article No. 4, Photo YCC Color Encoding and Compression Schemes <http://www.kodak.com/daiHome/techInfo/pcd-045.shtml> (Eastman Kodak Company 1998).
18. The first issue of RLG DigiNews (April 15, 1997, vol. 1, no. 1, <http://www.rlg.org/preserv/diginews/diginews1.html>) includes an FAQ section on identifying other digital imaging projects.
19. Clearing House of Image Databases and IMAGELIB listserv archives: University of Arizona Library<http://dizzy.library.arizona.edu/images/image_projects.html>undated.
20. Photo CD on the Web: <http://www.kodak.com/digitalImaging /cyberScene /cybersceneHome.shtml>.
21. More information on the FlashPix format, including retrieval software tips, is available at: <http://www.kodak.com/daiHome/flashPix/>.
22. The Judge II color-viewing booth provides metered daylight and balanced light levels to provide a reliable and controlled viewing environment. It is used for matching colors, evaluating color quality, and appraising color uniformity under specified lighting conditions. It can be purchased from GretagMacbeth (1-800-MACBETH)
23. Natural-daylight fluorescent bulbs can be purchased from lighting supply stores. When you are purchasing the bulb, be prepared to specify length (usually 2 feet to 4 feet) and color temperature (D5000 for daylight).
Association for Information and Image Management International (ANSI/AIIM). 1993. Resolution as it Relates to Photographic and Electronic Imaging. ANSI/AIIM Technical Report TR26-1993.
Allen, David Y. 1997. "Digital Imaging for the Rest of Us: Kodak Photo CD and Kodak Pro Photo CD." Meridian 12: 12-15.
Cartolano, Robert, Janet Gertz, and Susan Klimley. 1996. Oversize Color Images Project Phase II: Final Report to the Commission on Preservation and Access. Columbia University Libraries and Academic Information System (November 1996). <http://www.columbia.edu/dlc/nysmb /reports/phase2.html>.
Eastman Kodak Company. 1998. Kodak Digital Science, Photo CD Products and Information Web page. <http://www.kodak.com/daiHome /products/photoCD.shtml>.
- - - . 1998. Fully Utilizing Photo CD Images: Article No. 4, Photo YCC Color Encoding and Compression Schemes. Photo CD Information Bulletin Web page. <http://www.kodak.com/daiHome/techInfo/pcd-045.shtml>
Ester, Michael. 1994. "Digital Images in the Context of Visual Collections and Scholarship." Visual Resources 10:11-24.
- - - . 1995. "Specifics of Imaging Practice," p. 150 in Multimedia Computing and Museums. Pittsburgh, Penna.: Archives and Museum Informatics.
- - - . 1996. Digital Image Collections: Issues and Practice. Washington, D.C.: The Commission on Preservation and Access.
Frey, Franziska. 1997. "Digital Imaging for Photographic Collections: Foundations for Technical Standards." RLG DigiNews 1, no. 3 (December 15, 1997). <http://www.rlg.org /preserv/diginews/diginews3.html>.
Goseny, Michael. 1995. The Official Photo CD Handbook; A Verbum Interactive Guide. Berkeley: Peachpit Press.
GretagMacbeth. 1997. Fundamentals of Color and Appearance. New Windsor, N.Y.: GretagMacbeth.
Johns Hopkins University. 1998. Pathology Photography: Kodak Photo CD Web site. <http://photography.jhu.edu/PhotoCD.htm>.
Kenney, Anne R., and Chapman, Stephen. 1995. Tutorial: Digital Resolution Requirements for Replacing Text-Based Material: Methods for Benchmarking Image Quality. Washington, D.C.: The Commission on Preservation and Access.
- - - . 1996. Digital Imaging for Libraries and Archives. Ithaca, N.Y.: Cornell University Library.
Larish, John J. 1994. Photo CD; Quality Photographs at Your Fingertips. Torrance, Calif.: Micro Publishing Press, Incorporated.
Library of Congress. 1997. Request for Proposals for Digital Images of Pictorial Materials. Washington, D.C.: National Digital Library Program. RFP 97-9.
Reils, Charles. 1997. "It is Real Now: Web Page Color Matching on Remote Monitor Screens." Advanced Imaging 12, no. 9 (September 1997): 76-78.
Sandore, Beth, and Robert Dunkelberger. 1996. The University of Illinois/Kodak Digital Imaging Project: A Report of the Use of the Photo Imaging Workstation and Related Imaging Projects. University of Illinois at Urbana-Champaign Library. <http://images.grainger.uiuc.edu/KODAK.HTM>.
Technical Review. 1997. RLG DigiNews 1, no. 3 (December 15, 1997). <http://www.rlg.org/preserv/diginews/diginews3.html>.
Von Bulow, Heinz. 1994. The Photo CD Book. Grand Rapids, Mich.: Abacus.
Williams, Don. 1998. "What is an MTF . . . and Why Should You Care?" RLG DigiNews 2, no. 1 (February 15, 1998). <http://www.rlg.org/preserv /diginews/diginews21.html>.
- Mary Arsenault
- Department of Preservation and Conservation
- 214 Olin Library
- Cornell University
- Ithaca, NY 14853