Oversize Color Images: Addressing issues of Preservation and Access
oversized images

Oversized Color Images: Addressing Issues of Preservation and Access

Principal investigators

This project is made possible through a contract from the Commission on Preservation and Access and funding from Columbia University.

Starting Points

| Summary

| Rationale

| Goals

| Description Map Originals

| Selection of Vendors

| Description of Vendor Products

| Definitions for Evaluating Quality

| Results

| User Evaluation

| Bibliographic Control

| Costs

| Working with Vendors

| Organizational Issues

| Next Steps

| Appendix A: Chart of Effective Resolution

| Appendix B: Demonstrations of the Project


This report contains the results of Phase I of the Oversized Color Images Project to identify acceptable preservation and digital access techniques for dealing with oversize, color images associated with text. The goal of Phase I was to provide a preservation-quality photographic archival copy, a digital version for on-line access, and paper printouts. Five maps from brittle volumes were scanned. In addition, single-frame color microfiche (produced during an earlier project funded by the Commission on Preservation and Access) and 4"x 5" color transparencies, were scanned. Printouts were made from many of the digital versions.

Image quality, in terms both of capture and of display, involves a number of complex factors which are not always easy to consider and evaluate separately: characteristics of the original media, characteristics of the photographic intermediaries and how they were produced, scanning and display resolution, compression, storage, display hardware, display software, and network transmission limitations. This report concentrates primarily on resolution and on display issues relevant to modern printed maps, which consist of text superimposed on images made up of a relatively small range of colors without subtleties of shading.

Evaluation of the digital files and the highest quality printouts shows that fine details can be captured successfully from all three media (paper original, microfiche, transparency). For the purposes of creating an archival-quality storage medium combined with the potential for increasingly sophisticated electronic presentation of such materials, the project has been successful.

At this time the ability to capture information outstrips capacity for easy access and display with average equipment, so that immediate on-line use of high resolution files is somewhat limited. Because all printed details visible in the paper originals can be fully captured by current technology, however, re-scanning should not be needed in the future even though access and display capacity is expected to increase rapidly in the near future.

Rationale: Problems of Preservation and Access Presented by Oversized Color Materials

The Commission on Preservation and Access paper Scholarly Resources in Art History(1) and the report of its Joint Task Force on Text and Image, Preserving the Illustrated Text(2), have documented the inadequacy of traditional microform techniques to preserve brittle books characterized by a complex combination of text and illustrated materials. Such materials are relatively prevalent, especially in the literature of certain disciplines. The University of Chicago Library Preservation Department's 1989 Report on the Condition of the Collections(3) indicates that between 10-20% of their collections of general science, astronomy, natural science, agriculture, technology, and history of science contain oversize materials folded into bound volumes. A survey in 1994 by Columbia of architectural history volumes found that 68% contain halftone, continuous tone, or color illustrations; 17% contain oversize foldouts. Geography and the geological sciences are likewise characterized by heavy use of maps and other oversized illustrations.

Brittle volumes with oversized images suffer particular physical stress because the juxtaposition of text and images was achieved through folding oversized items into the binding or pockets. While the pages of the volumes follow the usual slow course of chemical degradation, self-destruction of the oversized foldouts is dramatically hastened as every folded edge breaks and information is irretrievably lost.

Maps present a particular challenge. Scholars need to be able to see the map as a whole, to follow information across its surface, and also to read fine details at every point. They have frequent need for paper copies to carry away from the desk. Color and pattern are both important as coding devices. Any reproduction must distinguish among colors which are distinct in the original, and distinguish among patches of the same color printed in distinct patterns such as parallel, horizontal or vertical lines.

While high contrast black and white microfilm serves well as an archival medium, it is inadequate to (1) convey color, (2) capture fine details of large images, (3) maintain the juxtaposition of image and text, and (4) provide methods for printing out full-size color copies of the images. When a map is filmed as a whole, the details are lost because the reduction ratio must be very high in order to fit the map into a 35mm frame. In order to capture the details, a map must be filmed in sections which can only be viewed sequentially. Printouts are often illegible, and in any case must be pieced together to recreate the entire map.

Concern over these issues resulted in the Commission's 1991/92 sponsorship of a project designed to experiment with preserving twenty-eight brittle numbers of the New York State Museum Bulletin through creation of single- frame color microfiche (4) and black-and-white scanned images. (Single-frame color microfiche use the entire surface of the microfiche for a single image, thus allowing for a significantly lower reduction ratio than microfilm, so that details can be read without sectioning the map.) Paper printouts were made from both versions. The sampled numbers, in subjects ranging from botany to entomology to the earth sciences, contained a wide variety of problematic and often oversized illustrative material including color drawings and maps. Although much was learned from that experiment of two years ago, one of the most tenacious problems proved to be the oversized color images.

Identification of a group of suitable preservation and access mechanisms for oversized images has remained an essential first step in preserving the literatures which utilize such images. Digitization offers us a wide new range of access options. A bitmapped image of a map captured at high resolution and with 24-bit color is a detailed copy of the map which can be saved in lossless TIFF format (Tagged Image File Format) as an archival record. A wide variety of use products can then be derived using JPEG (Joint Photographic Expert Group) compression or other space saving but lossy compression methods.

At the same time, the desire for long-term preservation of the information coupled with uncertainty about long-term digital refreshment and migration mean that we continue to want both a film master copy and a digital version. Most of these books and their oversized illustrations are brittle to the point of crumbling and we will not be able to keep the originals; we cannot yet feel confident enough to rely solely on the digital copy for long-term preservation.


The goal of the present project was to devise a methodology for providing both a film version of the image to serve as the permanent archival copy, and a digital version to serve as the access copy. Much of the cost of reformatting, whether by microfilming or scanning, lies in selection, cataloging, and preparation of the items and in actually handling the item during capture. Ideally the information should be captured directly from the original image only once and then transferred to other formats as desired. Since in this case the originals are not only oversized but also brittle, it would put less strain on the fragile originals if a film version could be made first and then scanned.

Among the questions we addressed in this project are:

To answer these questions, the project compared the results of scanning five oversize, color brittle maps using a variety of methodologies to produce both onscreen versions and paper printouts. Geologists, geoscientists, other subject specialists, and librarians and other interested parties offered reactions to the digital and paper versions.

Description of Map Originals

Five brittle oversized color maps from the New York State Museum Bulletin were selected for the project and borrowed from the New York State Library. These are the same copies of the maps which were used in the 1992 project described above. The size of the maps varies from 9"x 12" to 19"x 24"; thus, while they are larger than normal page size, they are relatively small for maps.

The project worked from three formats and compared the results of scanning each version.

Original paper maps

  1. Geological Map of Perce and Vicinity
    NYS Museum Bulletin 80 (1905)
    9"h x 12.5"w
    Hand-drawn cross section with watercolor-like pastel pigmentation.

  2. Part of the Catskill Quadrangle
    NYS Museum Bulletin 84 (1905), plate 7
    14.75"h x 9"w
    Three-color geologic map with contour lines in light brown, water in blue, and details in shades of pink.

  3. Surface Geology of Mooers Quadrangle
    NYS Museum Bulletin 84 (1905), loose map
    24"h x 17.5"w
    Typical full-color geologic map with fine detail and subtle color coding.

  4. Preliminary Geological Map of Portions of Herkimer... and Adjacent Counties
    NYS Museum Bulletin 95 (1905), plate 14
    8.75"h x 14.5"w
    Geologic map with coding in one color (light brown) in a variety of patterns and intensities.

  5. Geological Map of Buffalo Quadrangle
    NYS Museum Bulletin 99 (1906), loose map
    21.5" h x 19"w
    Typical full-color geologic map with fine detail including details of Buffalo city proper and cross section printed perpendicular to map along right edge.

The smallest type present on each map is 1 mm. Often this is the contour line elevation, which in the most challenging cases is printed in light brown in italics on top of color-coded areas. The result is very little contrast between the number and the background, and the numbers are difficult to read even in the paper originals. Even smaller are .6 mm squares representing buildings on the two larger maps, although these are in black and therefore easier to see.

The number of colors and their use and distribution on the maps varies. The simplest map uses only one color, printed in five codes which differ in pattern and intensity. The most complex map uses each of five distinct colors in three to four different patterns and intensities, for a total of seventeen different codes. The differences between the codes are often subtle and it can require close examination of the original to distinguish among them.

Single-frame color microfiche

The State Library's copies of the five maps had been used in the 1991/92 project by the German company Hermann and Kramer to produce single-frame color 105 mm microfiche on Cibachrome stock. A second generation set of those microfiche belongs to Columbia University and was used in the current project. The microfiche are 5.75"x 4.13"; the actual map images ranged from 3.25"x 3.38" to 3"x 4.88". Three vendors scanned these fiche.

One vendor produced its own set of single-frame microfiche from the same paper originals. Their fiche are also on Cibachrome stock, but are first generation. The size of the images differs. The maps of Perce, Catskill, and Herkimer range from 2.5"x 1.75" to 2.88"x 1.5", while the Mooers and Buffalo maps are approximately 4.75"x 3.5".


Color transparencies (4"x 5") were made from the same paper maps in the Columbia University Libraries Reprography Laboratory for the purposes of this project. Standard color charts were included with the maps. The film stock was Ektachrome 100 Plus.

Selection of Vendors

A request for proposal was prepared and sent to a number of potential vendors in order to determine which would be interested in participating in the project, and to see what imaging services might be available. The choice of vendors to whom we sent the RFP was based on previous contacts through other projects, the vendor fair held during the Research Libraries Group (RLG) Digital Imaging Technology for Preservation Symposium (March 17-18, 1994), and recommendations from colleagues.

Our RFP sought vendors capable of:

Vendors were required to:
Eventually five vendors responded positively to our RFP and took part in the project. They were:

Boston Photo Imaging, Boston, MA
Luna Imaging, Inc., Venice, California
Micro-Color International Inc., Midland Park, NJ
Stokes Imaging Services, Austin, TX
Visual Information Inc., Denver, CO

The maps, microfiche, and transparencies were sent to each vendor in turn. No one vendor was able to offer all of the services listed above, and each used somewhat different scanning and printing methodologies.

In scanning the originals, the vendors were asked to include standard color charts to allow for evaluation of the color quality displayed at the screen and printed out in paper versions. The second generation microfiche (produced during the 1991/92 project) unfortunately had not been shot with color charts.

Description of Vendor Products

The following describes the work performed by each vendor, the equipment used, the total number of vertical times horizontal pixels used for the image (not to be confused with dots per inch), compression method, and comments. All of the scanned images are available over the Internet. The set numbers in the descriptions herein refer to the on-line images accessible on Columbia's World Wide Web server URL http://www.columbia.edu/dlc/nysmb/

Boston Photo (set 1 in the on-line version)

Boston Photo worked only with the transparencies, and did not produce paper printouts.

set 1a.Produced 35 mm slides from the transparencies and scanned them on a Kodak PCD 2000 scanner to Kodak Photo CD at a maximum resolution of 2048 x 3072 (vertical x horizontal pixels). Because of the loss of quality in creating a derivative slide, this set of images was not considered successful.

set 1b. Scanned the transparencies directly on a Kodak PCD 4045 scanner to Kodak Pro Photo CD at a maximum resolution of 4096 x 6144 pixels and 24-bit color.

The compression method used in both cases was the proprietary Kodak Photo CD format.

Micro-Color (set 2 in the on-line version)

set 2a. Scanned directly from paper originals using the Leaf Digital Studio camera at a resolution of 2048 x 2048 pixels and 24-bit color, wrote the files to Kodak PhotoCD and also stored them as uncompressed TIFF files.

set 2b. Created new first generation microfiche and scanned them (not the 1991/92 fiche) using an Optronics drum scanner at resolution of 2000 dpi (ca. 4000 x 5000 pixels for the two larger maps and 2500 x 3500 pixels for the three smaller maps) and 24-bit color. The files were written to Kodak PhotoCD and also stored as uncompressed TIFF files.

The compression method used in both cases was the proprietary Kodak PhotoCD format.

2c. Printed out from both using Iris Smartjet 4012 for the three smaller maps and the Iris 3024 printer for the two larger maps. All were printed on coated paper

Luna (set 3 in the on-line version)

set 3a. Scanned from transparencies to Kodak Pro PhotoCD using resolution of 4096 x 6144 pixels and 24-bit color, and Kodak PhotoCD compression. They also produced uncompressed RGB TIFF files. They scanned the Q60 standard color chart and used it to color-correct the images based on color bars in the transparencies.

set 3b. Scanned from second generation 1991/92 microfiche to Kodak Pro PhotoCD using a resolution of 4096 x 6144 pixels and 24-bit color, and Kodak PhotoCD compression. They also produced uncompressed RGB TIFF files. They attempted color-correction of the images made from the microfiche by correlating to the color of the images made from the transparencies, where color bars were included.

Luna used the Kodak 4200 Professional Imaging System, consisting of:

Luna also provided a diskette listing image numbers, conversion types and date, resolution and bit-depth, rotation, nature of the item being scanned, and related information.

3c. Printed out from both sets of images onto matte paper using an Iris 3047 printer.

3d. Printed out 11"x 11" prints from both sets of images using the Kodak Photo Printer XL7720 onto Kodak electronic print paper.

Stokes ( set 4a in the on-line version)

Stokes intended to use a digital camera to scan the original maps in sections at high resolution and then concatenate the sections to produce one large image ("tiling"). This plan was based on their expectation of acquiring appropriate hardware and software in time for the project. By that time they were not yet equipped to concatenate the sections and handle the huge files which would have resulted. Instead, Stokes scanned each original map in three ways.

First, the entire map (up to 14.63"x 19.63") was scanned at 2984 x 2312 pixels using a Kontron digital camera. Second, a 9"x 12" section was scanned at the same resolution, and finally a 6"x 12" section at the same resolution. As the section grows smaller, the effective resolution grows higher; for instance, the resolution of the entire Catskill map is 204 dpi, the 9"x 12" section is 270 dpi, and the 6"x 8" section is 373 dpi.

JPEG compression was used. It was written at a quality factor of -5 (compared to the highest quality at -7).

The second generation (1991/92) microfiche and transparencies were sampled to produce scans of the maps and the sections which matched the scans from the original in dpi. The 6"x 8" section of the original is captured on a 1"x 1.25" area of the fiche (and also the transparency). That area was scanned at 2387, which translated into a resolution of 373 dpi on the original. In other words, the scan of the 6"x 8" section of the original, the 1"x 1.25" area of the fiche, and the 1"x 1.25" area of the transparency when viewed side by side are all the same size and the same resolution.

The sectioning does allow for much higher resolution views of the maps. Unfortunately, a number of the images are somewhat out of focus along the right side. This appears to be something that could be corrected through rescanning.

Visual Information (set 5a)

Scanned the original maps. Both Tangent and Sharp flatbed scanners were used. Smaller maps were scanned at 300 dpi and larger ones both at 300 and 400 dpi. All scanning was done at 24-bit color.

Wrote the images of the three smaller maps to a CD-ROM containing their proprietary ImageBASE TM software; and wrote the images of the two larger maps to 8mm tape. The files were also provided in uncompressed TIFF format.

Printed out the Buffalo map from a reduced resolution image using a Hewlett-Packard color printer, onto matte paper which was then laminated.

Definitions for Evaluating Quality

In evaluating images the question arises how to define success. In this project we were not seeking to achieve the best possible quality image, but rather, the image that is good enough to serve fully the needs of researchers in lieu of the original paper maps. In order to do this, a reproduction of the original map must reproduce all details and must reproduce the color codes accurately. Quality control was performed on a Tektronix 19" 1280x1024 X terminal to assure consistency, but images were also viewed on a variety of other equipment.

Our definition of adequate quality has two parts:

The definition of fully successful quality is:


In looking at the products delivered by the vendors, it is clear that it is possible to achieve adequate results starting from the original map, the microfiche, or the transparency. The 1 mm elevation marks are legible at high resolution, and all color pattern codes can be distinguished. There has been some color shift in all printouts and in all images as viewed at a variety of monitors, however.

These results mean that for preservation purposes we can safely continue to create preservation-quality microfiche and transparencies to serve as surrogates or replacements for brittle originals. We can be confident that we can later scan the film intermediaries even if the originals have been lost, and still capture the intellectual content of the originals.

It must be emphasized that we have dealt here only with modern printed maps and that our goal has been to capture the printed content, which in these materials consists of text superimposed on a relatively small range of colors without subtleties of shading. Quite different results would be expected for archival materials, reproductions of works of art, and other images with a wider palette and more subtle detail.

It should also be noted that some vendors were more successful than others at producing digital images that met our definitions of adequate quality. For instance, although most of the vendors provided digital images with resolution of at least 150 dpi or higher, in some cases the images were in better focus than others. Variations can be due to the particular scanners used, operator skill, and similar factors. In the same way, some vendors images have better color quality than others, thanks to a combination of quality of lighting and care in balancing color and calibrating equipment.

Effective Resolution:

Effective resolution refers to how many dots per inch the digital file provides relative to the paper original, regardless of whether the image was made from the paper original, the microfiche, or the transparencies. Thus, if an image consists of 2000 x 2000 pixels, and it represents a scan of a 10"x 10" original, the effective resolution is 200 dots per inch (2000 pixels divided by 10 inches). Calculation of effective dpi includes the space devoted to color bars. If the same 10" original were accompanied by a color bar which is 3" thick and placed 1" from the map, the width would be 14" and the dpi would be 143 (2000 pixels divided by 14 inches).

When discussing the resolution needed to capture the same degree of detail from a film intermediary as from the original, effective resolution is determined relative to the original: the film version must be scanned so as to provide the same number of pixels across the face of the miniature image as were used in scanning the full-size original. Assuming that the 10" original printed map has been filmed as a 4" image on a microfiche, each inch of the microfiche represents 2.5 inches of the original. Since the 10" original requires 2000 pixels to achieve an effective resolution of 200 dpi, the microfiche of that map will also require 2000 pixels. But the 2000 pixels which stretched across 10" of the paper map must now be fitted across the 4 of the microfiche, meaning that the microfilm is scanned at 500 dpi (2000 pixels divided over 4" inches).

The effective resolution has been calculated for each project image and reported in Appendix A.

Achieving successful resolution depends on a combination of the size of the map, the size and color of the smallest type, and the complexity and range of the color coding. Lower resolution (normally less expensive to attain) will suffice for smaller and simpler maps. This is clear in comparing the map of Herkimer with the map of Mooers Quadrangle.

The 13.5"x9" map of Herkimer is very simple. It has five codes in one color (light brown) that differ by pattern and intensity and are easily distinguished from each other. The smallest type is 1.2 mm; all type is printed in black and quite easy to read in the original.

In set 3B (made from the second generation microfiche, without color bars), the smallest type on the Herkimer map is just barely legible on screen at a resolution of 1024 x 1536 pixels and the five patterns are distinct even though the fine parallel horizontal lines of the first code cannot be distinguished and the area appears to be a solid color. At this resolution the map is being viewed at approximately 66 dpi. At 2048 x 3072 vertical x horizontal pixels all the fine details of the map are fully visible. The effective resolution is 132 dpi and the display is actually larger than the original map. In set 5B at 2346 x 1465 pixels the map is being viewed at 174 dpi and it looks extremely good.

The Mooers Quadrangle map, in contrast, is larger and more complex. It is 17.5"x 24", a typical full-color geologic map with fine detail and multiple color coding. It uses fourteen different codes in varying patterns and intensities of orange, green, pink, beige, and gray. Contour lines with elevations printed in 1 mm beige type are superimposed on top of the color coding. They are difficult to read on the original map without a magnifying glass.

In set 3B, at a resolution of 2048 x 3072 pixels, the smallest type on the Mooers map is not legible on screen, and it is just barely possible to read the labels in the key to the patterns. There is some difficulty distinguishing among the fourteen code patterns. The map is being viewed at approximately 87 dpi. At 4096 x 6144 pixels the contour elevations can be read and all codes can be fully distinguished. At 4096 x 6144 it is being viewed at approximately 174 dpi effective resolution.

Due to the difference in size, the Mooers map must be viewed at double the resolution to reach about the same dpi as the Herkimer map. It appears that something in the range of 130-170 dots per inch on the paper original may be sufficient for both maps to be usable although not for the finest details to be legible. The formula cannot be based simply on the size of the map and the size of the smallest type, however. The fact that the Mooers map is visually also much more complex, not merely that it has smaller type than the Herkimer map, means that Mooers requires higher resolution to make the finer details fully legible, especially where the lack of contrast between the elevations and the background plays a role. If Mooers were as simple as Herkimer, a somewhat lower resolution might suffice.

While some urge capturing all images at the technically highest possible resolution, the results of this project indicate that full information can be captured at lower resolutions depending on the size of the smallest type and on the pixel depth. In this case where we are using 24-bit color and capturing a 1 mm high letter, we need 200 dpi for full legibility. Someone scanning 1 mm high black and white text in binary mode would need 600 dpi for equivalent legibility. In either case, once the full information is captured at the highest resolution required for full capture, it should not be necessary to re-scan in the future, since no information would be added by scanning at the higher resolution.

These findings are in line with the proposals on required resolution which Cornell is putting forth for gray scale and color materials (6). They have proposed formulae for binary (2 bit) and grayscale (8 bit) scanning to derive benchmarks for recommended dpi capable of giving resolution equivalent to that of microfilm at quality index level 8 (the level required by the Research Libraries Group). Following the logic of their formulae, a formula for color scanning (24 bit) suggests 200 for quality index level 8.

Based on the results of this project, it appears that a resolution of approximately 200 dpi in scanning the original captures all distinctions made on the original in full clarity. For purely access purposes on the other hand, display at lower resolution in the 130-150 dpi range leaves the larger type legible and allows for browsing.

It should be noted that the file of the 4096 x 6144 pixel version (i.e., 200 dpi) of the original Mooers map is over 18 megabytes in size, so large that it is not easily accessed and manipulated by many computers. At 17.5"x 24", Mooers is not an enormously large map. The clear conclusion is that even larger maps with equally small print would require extremely large files in order to capture the print legibly. Tiling is one potential solution to this problem which was proposed by one vendor but not completed. It bears further investigation.


All scanning was done at 24 bit color, where every pixel includes 8 bits of information each for red, green, and blue. The combination of the 24 bits of information potentially produces 16 million different color combinations, allowing for great richness of hues and extreme accuracy of reproduction if the scanning equipment is properly calibrated.

In order to determine whether colors would shift as a result of scanning, we requested that scans of the originals include standard color and 18-step gray scale charts. The charts were also included when the transparencies were made. Color and gray scale bars were incorporated in the microfiche of the 1991/92 project, unfortunately.

As noted above, color had shifted in all screen versions and all printouts. Shifts might be expected in the case of the transparencies and microfiche, since change is introduced by every aspect of the photographic process: the specific film stock, lighting, developing, printing. Yet shifts also occurred in the images printed from the originals. Even in printouts from the same medium from the same vendor, differences among the color bars could be distinguished. It is at least theoretically possible, however, to calibrate scanners, monitors, and printers by defining each of the 16 million colors relative to a standard, as long as standard color charts are included in the process. As long as standard color charts are present, use of color manipulation software can assist in bringing the colors closer to the original. This should not be considered "enhancement" but rather a way to reduce the amount of interference in the capture of the original color. It should be noted that such manipulation takes time and therefore increases costs.

Given the nature of these maps, the information conveyed by the color codes is adequately captured as long as the codes remain distinct, even though the actual hues may be altered. When considering the preservation of an accurate copy of the original, however, more work is needed to assure that the color has been captured precisely; developing methods for accurate color capture, display, and printing is obviously a priority.

One complication in connection with the color charts should be noted here. Because we did not specify the placement of the color and grayscale bars, vendors used different arrangements. Some were more careful to maximize the use of space in the placement of the bars and orientation of the maps, while others left significant empty space around the maps and bars. Different scans of an individual map are therefore not fully comparable because the map may take up more or less of the image field at the same resolution depending on how much space is filled by color bars and margins. A map which fills the field is at a higher resolution than the same map when it fills only three-quarters of the field. While cropping of the blank margins is possible, it requires manual intervention.

In sum, color and grayscale bars are important for calibrating color and minimizing shifts, but they take up space and pixels. The smallest available bars should be used, and they should be placed so as to minimize waste of space. Likewise, when scanned the map should be oriented to fill the image field as fully as possible. In presenting the images to users it would be possible to crop the color charts to save space, but the color bars should certainly be kept intact in the "master" file and made available to users who want to evaluate and/or manipulate the color they receive on screen.

Paper Printouts:

The quality of the paper printouts from Iris machines provided by both Luna and Micro-Color is quite adequate. Fine details down to the smallest 1 mm numbers can be read on all of the printouts, and although colors shifted all coding remained distinct.

The Iris printouts created for this project are strictly for use purposes, and do not meet preservation criteria. The two types of paper used, a light-weight matte paper and a heavier glossy paper, are not acid-free. The pH averages 5.3-5.8. The inks are water soluble and there is no information available about how long the inks last before fading. A comment from Micro-Color indicates that the manufacturer may be working on dyes with greater stability, but no details were available. It would also be interesting to see whether the printers can accept acid-free papers.

The 11"x 11" prints made by the Kodak Photo Printer XL7720 on Kodak electronic print paper are surprisingly legible, even though in the case of the two large maps the prints are less than a quarter of the size of the originals. Under a magnifying glass the finer details remain readable. Again, the paper is not preservation quality (pH is 5.5) and the inks are not stable; however, they are a very viable use medium.

Finally, a number of printouts were produced on various desktop printers. The Apple Color Style Writer 2400 printouts were reasonably legible, although limited to 8.5"x 11" sections of the maps. Other lower-end printers were able to provide enough detail to suffice for browsing or "shorthand" copies.

User Evaluation

Evaluation of the vendors' products was carried out on several levels. At Columbia the project team, other librarians and computer experts, and faculty of the Department of Geological Sciences evaluated both the on-line images and the printouts. The on-line images and the paper printouts were shown away from Columbia several times and audience reactions recorded. Demonstrations were presented to attendees of the Geological Society of America (GSA) conference in October, at which the Geoscience Information Society (GIS) also meets. A report of the demonstrations is attached, as is a report of a second demonstration presented at the New York State Library, attended by a variety of library and geology specialists. See Appendix B: Demonstrations of the Project. A description of the project was also presented to the Reformatting Committee of the Preservation and Reformatting Section of the American Library Association. Other demonstrations are planned for the United States Geologic Survey. The images have also been made available on the Internet for any interested parties to view and send comments, and many have.

Files Accessed Over the Internet:

One goal of the project was to make the digital images of the maps available on the Internet to see if this is a viable way to give scholars access to the maps. A narrative description of the project accompanies the map images. The files are mounted on Columbia's World Wide Web server, where they can be accessed by anyone with Mosaic or Netscape capability at the URL:

The existence of the images was announced on a variety of listservs read by librarians, geologists, and others interested in digital imaging. The people who responded were in a variety of disciplines interested in images, including for instance an experienced network user in medical informatics.

As stated above, the high resolution scanning resulted in files of 18-24 megabytes or more. Such large files cannot be accessed and manipulated except with high-end equipment. Therefore, in order to make the files usable on-line, Columbia has provided versions at lower resolutions accessible to most users. The on-line display states the file's size for each image so that users can avoid trying to call up images which are too large for their equipment. For some images two versions are available, one with 256 colors in lossless GIF format, and the other with full 24-bit color but using lossy JPEG quality 50 compression. Examples of versions which have been made available on-line are:

vertical x horizontal pixelsapproximate file size
512 x 768 GIF200-250 KB
1024 x 1536 GIF1.5 MB
1024 x 1536 JPEG77 KB
2048 x 3072 PhotoCd5-8 MB

While many people found retrieving the maps easy, there were unfortunately -- but not unexpectedly -- some problems with remote access. One of the Columbia WWW servers was down most of the week that the URL was announced publicly, and this caused many people to be unable to access it or to have to try several times before succeeding. Even after this problem was solved, the usual facts of Internet life remain as barriers to consistent easy use of images.

Access through the WWW takes time. Even on the Columbia campus when traffic is heavy it can take several minutes to load the large file of a high resolution image, and accessing it remotely can take hours longer, especially when the Internet is busy. The smaller, lower resolution files move faster, but cannot provide all the details. An obvious solution would be to determine what the lowest resolution is which is successful as a browsing tool and mount that as the file recommended for real time access, with high resolution versions available for downloading during low-use periods.

All of the evaluators were asked to determine which images (both scanned and printed out) were acceptable/not acceptable for identification and research purposes. The responses to the various images were overwhelmingly positive. Evaluators were generally enthusiastic about the quality of resolution, which made all but the largest and most complicated map (which also has some of the smallest print) fully legible on-line. There was agreement that even the lowest resolution image of the map of Herkimer at 512 x 768 was a total success for browsing purposes when viewed on a variety of monitors.

Viewers observed that the more complex and larger maps could not even be browsed at low resolution; at the same time the higher resolution images of the large maps could not be viewed as a whole on the screen. Inability to view the whole surface can sometimes be disorienting when trying to read a map, and studies have indicated that peripheral vision is very important in map use.(7) Several people suggested sectioning the maps as an intermediate option; others did not want to give up the ability to move around the whole map within one file even if only part of it was visible at one time. Viewers concurred that although we might not be able to view the whole large images today, it was just a matter of time before desktop computing capacity caught up.

Other aspects of on-line use brought less favorable comments. Viewers were concerned about access difficulties and the time and memory size needed to view the maps. While many in the geological sciences are not interested in the quality of the color (as long as they can distinguish easily among the codes), a few commented on the color shifts as an issue that should be addressed. There was some discussion about scale. Most maps carry a scale indicating what distance is represented by an inch or centimeter, and some evaluators felt it would be helpful to add a note of reminder to viewers that the size of the map was not necessarily one to one on the screen or in a printout. Whether or not a scale was explicitly printed on the map, evaluators stated that it would be very important to include a ruler in the image so that the original size could be easily determined (since the color bars include a ruler, they thus can serve two purposes). A final issue was the degree of distortion which may have been introduced through scanning. Use of lenses in photography is known to distort the image's edges and investigation was urged into whether scanning directly from the original might eliminate the potential for distortion.


Evaluators viewed the three sets of Iris prints as well as the 11"x 11" photo prints, and a sampling of 8.5"x 11" printouts from a standard desktop color printer. They were generally quite happy with the Iris and the 11"x 11" prints as use media. While the small prints were seen as less useful than the full-size printouts for serious research, a number of evaluators considered them a handy and inexpensive intermediate product that could fill needs for identification and browsing. Evaluators were also happy with the option for local desktop printing. Although desktops produce only 8.5"x 11" sections of the maps and have limited color ranges, the printouts provide a very convenient sketch of the map for identification purposes for anyone who does not need the full color, full size version.

Evaluators were divided in their preference for the matte versus the glossy paper, and agreed that it would be good to have both types. They noted that some of the Iris prints were not sized exactly to the originals and urged that this be corrected in the future. Again, printing the map with a ruler to the side would help in verifying the scale used in the map. Several also noted the color shifts and hoped that this could be corrected in the future as well.

Printout longevity was a concern. Many evaluators said that it would be highly desirable to be able to produce printouts on demand equivalent in permanence to preservation photocopies with which to replace out-of-print damaged and brittle originals. Several asked how soon Columbia would be able to make copies available on demand and indicated immediate interest in purchasing sets of the maps.

Bibliographic Control

As part of the project we also decided to experiment with bibliographic control for the digitized maps. While the Internet allows wide access to electronic resources, it lacks any coherent type of indexing to let people know what resources are available. Since the national databases serve this function well, we feel it is important to add MARC records for digital files to RLIN/OCLC in order to alert researchers to their existence.

Guidelines and standards for creation of records for digital copies of paper materials are far from established, and Columbia is working actively to develop them with RLG, OCLC (through its Internet Cataloging Project), and the Intellectual Access Committee of the American Library Association's Preservation and Reformatting Section. Issues still to be decided include what data about images should be recorded and where -- what portions of this information might be part of the bibliographic record, which recorded with the image elsewhere. Useful/necessary information includes: the type of film stock used for photographic versions which have been scanned; the reduction ratio; and the dimensions of those photographic versions. Further data concerns the method of scanning, any color correction or other manipulation, equipment and software used, storage media, and so forth.

In the meantime, for this project we have created a collection-level bibliographic record for the project as a whole. The digitized images are described and subject access is provided. The record also refers people to the Web site for access to the files and the narrative report on the project. A 5xx note provides a narrative reference, and the URL is also entered into the 856 subfield where eventually it will constitute a hyperlink.


One area that needs much more attention is the basis for predicting costs. Because this was such a small project, several vendors donated their services and those that charged for the work did not have a production-level basis on which to estimate realistic costs for services they do not normally provide. In any case, the pricing of digitization services and equipment is constantly in flux. As a result, it is not possible to draw conclusions about relative prices from this project. We can simply report the range of prices charged:
scanning the original maps$75 each
scanning the microfiche$25-32 each
scanning the transparencies$28 each
scanning the transparencies$28 each
Iris prints of 2 larger maps$145-250 each
Iris prints of 3 smaller maps$40-55 each
11"x 11" prints$8 each

There were also a variety of separate charges for writing files to the storage media (PhotoCD, tape, etc.) that ranged from $4-20 per file.

Clearly there is a need for cost models for planning in a full production level digital project. Yale and Cornell's work on what it takes to "ramp up" will be very important in trying to get an idea of what digital imaging really costs.

Working with Vendors

This project has made clear the need for explicit communication between the library and its vendors. The documentation provided by the vendor is a key part of a successful project. Otherwise, identification of files and determination of exactly what resolution, compression, etc., were used becomes extremely difficult. Some vendors voluntarily provide detailed documentation; in other cases the library will need to specify what information it wants from the vendor, and in what form. The information contained in the file header is not sufficient. Questions of how to record data about images and intermediaries have begun to be discussed by libraries, and once a common set of guidelines is established it will be much easier to specify for vendors what information is needed and in what form.

Simply conveying to vendors what services and product quality are desired is not always easy. At present we cannot point to a convenient accepted standard but must spell out in detail what we want. Communication is hampered by a lack of shared vocabulary and clear standardized language to describe what we need. There is also a lack of understanding by librarians as well as some of the vendors as to what some technical terms mean, complicated by the rapid pace of change industry-wide.

Because quality control procedures for vendors and especially for the library are not yet established, it is not always clear how to guarantee that we are getting what we asked for from vendors. It can also be difficult to negotiate re-work with a vendor if we are unsatisfied with the product that results.

The solution is partly a matter of educating ourselves to become literate in digital imaging; and partly a matter of finding vendors who are interested in working with us. As was (and is) the case with microfilming, most vendors have no idea what preservation is and what our goals are. Some vendors are more concerned with selling the product they already produce rather than tailoring their operations to our needs. We must continue to identify vendors who are interested in learning what we want and are willing to work with us to develop preservation-quality services.

Organizational Issues

Digital projects are extremely complex to design and manage. They require expertise not normally found in a single library department or division, and sometimes not found in the library at all. Such projects are beyond what a single curator or a single preservation department can afford to tackle alone. Issues such as decisions on institutional policies, commitment of technical and hardware support, provision of financial support, planning for the future, assurance that digital files will be maintained, refreshed, and migrated, must be made on a system-wide basis.

Columbia uses a team approach to digital library issues. A standing committee has been established by the Libraries to oversee the progress of the digital library, coordinate projects, and to discuss issues in depth and come up with decisions. The committee is headed by the Deputy University Librarian and is composed of: the Director for Library Systems; Academic Information Systems (formerly Academic Computing) staff responsible for coordinating faculty and library imaging projects and mounting them on the Columbia Web server; the Director for Bibliographic Control; curators of collections in which projects are currently taking place or who have a high degree of experience with digital issues; and the Director for Preservation.

Each individual project also has a management team that brings together expertise on the subject matter and the needs of users, preservation, and the technical side of imaging. This particular project has a three-member team from Preservation, Geology, and Academic Information Systems. We have found that the team approach works very well and recommend it to other institutions.

Next Steps

Phase II of the project, an attempt to create an integrated on-line version combining the text and the illustrations, is now underway. The project is assembling a full film copy of four volumes of the Museum Bulletin, using a combination of microfilm of the text and single-frame color microfiche of the illustrations; and then scanning the microfilm of the text at 600 dpi black and white; and scanning the microfiche of the illustrations at approximately 200 dpi, 24 bit color. This will produce a series of digital files, one for each page and for each illustration. The project will use indexing and document structure software to integrate the files of pages and illustrations so that users can move easily from one to another, and recreate the original volumes virtually on the Internet.

The result of the project will be the full preservation of four sample volumes: long lasting microfilm/fiche will have been created for all the text and illustrations, and a truly usable digital version of them where the author's juxtaposition of words and illustrations is also preserved and available for users to read on-line or to create paper printouts.

A number of other areas have been identified as needing further research. It will be very important to determine what the real costs of digital imaging are, so that decisions on what endangered materials to convert and when can be made on a firm financial basis. Guidelines for adequate and high-level quality of resolution for various media and for various levels of use must be developed, along with schemes for institutional quality control evaluation to assure that required levels of quality have been provided by vendors and in-house scanning services. Improving the accuracy of color capture and display is another obvious area for research if digital imaging is to serve disciplines such as art history, where faithful representation of color is essential. Finally, the potential for tiling as a way to deal with very large images needs to be investigated. How difficult is it to achieve exact registration when moving from one section to the next? Would scholars find this a useful method for working with large images? How would tiling affect the role of peripheral vision in the use of large images?

The use of digital imaging for preservation purposes offers exciting possibilities. While much further research awaits, it is our hope that this project on the digitization of oversized color images will help to answer some of the questions.

Appendix A: Chart of Effective Resolutions

The dimensions of the map plus color bars and margins are given under each map name. Note that dimensions change from set to set because vendors placed color bars in different places and left different sizes of margins. Effective resolution has been calculated using these dimensions.

Most of the images in sets 4A-4C consist of sections rather than the entire maps; only the first group in set 4A contains images of the entire maps and only it is included here.

Set lB transparencies shot with color bars
display size in horizontal x vertical pixels
(23.25"x 12.25")
22 dpi44 dpi88 dpi176 dpi
(16.5"x 18.5")
31 dpi62 dpi 124 dpi248 dpi
(20.75"x 12.5")
25 dpi49 dpi 99 dpi197 dpi
(25"x 27.5")
20 dpi41 dpi 82 dpi164 dpi
(24"x 25")
21 dpi43 dpi 85 dpi171 dpi

Set 2A original maps shot with color bars
display size in horizontal x vertical pixels
512x7481024x15362048x3072 4096x6144
(16"x 14.75")
32 dpimissing missingnot provided
(17"x 15")
30 dpi60 dpi 120 dpinot provided
(17"x l5")
30 dpi60 dpi 120 dpinot provided
(25.5"x 24")
20 dpi40 dpi 80 dpinot provided
(25.75"x 21.5")
20 dpi40 dpi 80 dpinot provided

Set 2B first generation microfiche shot without color bars
display size in horizontal x vertical pixels
512x7481024x15362048x3072 4096x6144
(14"X 8.75")
37 dpi73 dpi 146 dpi293 dpi
(10.25"x 15")
50 dpi 100 dpi200 dpi400 dpi
(14.5"x 9")
35 dpi 71 dpi141 dpi282 dpi
(25"x 27")
not available not availablenot available not available
(20"x 25")
26 dpi51 dpi 102 dpi205 dpi

Set 3A transparencies shot with color bars
display size in horizontal x vertical pixels
512x7481024x15362048x3072 4096x6144
(23.25"x 12.25")
22 dpi44 dpi 88 dpi176 dpi
(16.5"x 18.25")
31 dpi62 dpi 124 dpi248 dpi
(20.75"x 12.5")
25 dpi49 dpi 99 dpi197 dpi
(25"x 27.5")
20 dpi41 dpi 82 dpi64 dpi
(23"x 25")
22 dpi45 dpi 89 dpi178 dpi

Set 3B -- second generation microfiche shot without color bars
display size in horizontal x vertical pixels
512x7481024x15362048x3072 4096x6144
(14.5" x 8.75")
35 dpi71 dpi 141 dpi282 dpi
(16.5"x 15")
45 dpi89 dpi 178 dpi356 dpi
(15.5"x 9")
33 dpi66 dpi 132 dpi264 dpi
(23.5"x 24")
22 dpi44 dpi 87 dpi174 dpi
(21.5"x 21.5")
33 dpi66 dpi 132 dpi264 dpi

Set 4A original maps shot with color bars
display size in horizontal x vertical pixels
(12.5"x 15")
119 dpi239 dpi
(14.5" x 11.25")
103 dpi 206 dpi
Mooers (margins trimmed)
(15.5"x 24")
96 dpi 192 dpi
Buffalo(margins trimmed)
(16"x 21.5")
93 dpi 187 dpi

Set 5A -- original maps shot with color bars
display size in horizontal x vertical pixels
(16.5"x 12.25")
99 dpi
(13.75"x 18.5")
119 dpi
(18.5"x 12.5")
87 dpi
(21.5"x 27.5")
76 dpi
(23"x 25.5")
71 dpi

Set 5B -- second generation microfiche shot without color bars
display size in horizontal x vertical pixels
(each maps display differs)
(12.5"x 8.75")
96 dpi
192 dpi
(9"x 15")
91 dpi
180 dpi
(13.5"x 9")
61 dpi
174 dpi
(21.5"x 27.5")
not available not available
(23"x 25.5")
not available not available

Appendix B: Demonstrations of the Project

Demonstration of Preliminary Results of Oversized Color Image Project at the Annual Meeting, Geological Society of America, Seattle, Washington, Oct. 23-26, 1994

At the invitation of the Geoscience Information Society (GIS), an affiliate society of the major geology professional society, Susan Klimley used the GIS booth in the main exhibit hall to demonstrate and get reactions to the preliminary results of the Oversized Color Image Project. The response was extremely positive to the paper and digital files that were shown.

Four images from the project were loaded onto a Macintosh PowerBook complete with a Mosaic interface. The front page of the project, which included the world wide wide locator (URL), was distributed to booth visitors. The images were arranged in a Quick Tour and illustrated the on-line success of the smaller images and the problems with on-line display of the larger images. It was originally planned to demonstrate the project accessing the URL over the Internet; however, it proved impossible to get needed telecommunications links. The PowerBook, hooked up to a large monitor, provided the look and feel of network access to the Oversized Color Image Project without the undependabilty of Internet connections.

The response of geologists and geology librarians was immediate and positive. The large monitor could be seen from some distance and many people were drawn to the project. A steady stream of conference attendees visited the booth listening to a brief description of the project and a comparison of the small and larger images. Just about every visitor examined the print products. Many visitors stayed to discuss technical aspects of the project, asking questions about how the scanning was done, the size of the files, printing, etc. Without exception people were happy with the results. There was agreement that the smaller image, the map of Herkimer at 512 x 768, was a total success viewed on the monitor. Visitors observed that the larger image, the map of the Catskill Quadrangle at 1024 x 1536, was barely viewable but concurred that although we might not be able to view the whole large images today, it was just a matter of time before desktop computing capacity caught up. Several people suggested sectioning the maps as an intermediate option, something that Rob Cartolano at Columbia had also suggested.

The printed maps were also received favorably. People were impressed with the high quality on the large map and the readability of the smaller map printed on a black and white laser printer. No one remarked on color variations among the printed copies. Several librarians asked about the stability of the paper on which the maps were printed, saying that it would be nice to replace heavily used and increasingly damaged maps in their collections with more archival copies. The slick surface of the map was commented upon and a number of people assumed the maps had been laminated. There was concern about the high cost of the maps.

People visiting the booth included academic and government geologists, geologists from state geological surveys, librarians, publishers and students. Several of the state geological survey geologists asked how they would get the publications of their surveys in the queue to be preserved in this manner(!). Many people asked what the next steps would be. Several students and geologists wanted a catalog of what was available in this format on-line. Over the course of the meeting, a number of librarians examined the process of accessing Mosaic and the Oversized Color Image Project URL carefully in preparation for using it in their home institutions.

A printed copy of the homepage which included the URL was available as a handout at the booth and most people took a copy. Visitors were encouraged to visit the site and fill out the questionnaire. Early in the conference the science librarians at the University of Washington tried to access the URL via a mosaic interface on the Internet and they were not successful. This gave rise to concern about whether the URL had been correctly recorded on the handout. The URL was verified but the situation pointed to the growing problem of heavy traffic on the Internet and increased inability to access operating sites. Librarians, who saw the demonstration at GSA, have reported mixed success in accessing the Oversized Image Project. It does raise concern about providing access to the results of this project much less to a large library of oversized color images.

As a result of the positive reaction to the Oversized Color Image Project, the Preservation Committee of the Geoscience Information Society requested that GIS contact the Geological Society of America to set up a joint Scholarly Advisory Committee on the Preservation of the Geology Literature. A representative of the American Geophysical Union expressed an interest in participating. GIS members indicated the importance of involving the American Geological Institute, a consortium of earth science professional organizations, in an interdisciplinary effort. Work on setting up the Scholarly Advisory Committee will continue as one of the GIS efforts for 1994/95.

Report prepared by Susan Klimley

Demonstration at the New York State Library, Albany, NY, November 22, 1994

James Lane, Preservation Officer of the New York State Library, arranged for Janet Gertz to demonstrate the Project as far as it had progressed by November 22, 1994. Attendees included staff from the State Library and the Archives, the State Geological Survey, the State Department of Transportation and the Department of Environmental Conservation, the State University of New York at Albany Preservation Department, the New York State Newspaper Project, the State Museum, and the United States Geological Survey. A handout provided a brief description of the project, the rationale for a hybrid film-plus-digital approach to preservation of color and oversized materials, and the project URL.

The digital versions were displayed in three ways. First, staff at the State Library had previously accessed and downloaded several of the available images via Mosaic. These were displayed (at a Pentium PC with a 17" monitor) without the delays of real-time access over the Internet. Second, one image was accessed real-time at about 10:30 am as a demonstration of the Mosaic commands (and the time delay that can ensue with remote access). The image took about 5 minutes to arrive. Third, four images from the project had been loaded at Columbia onto a Macintosh PowerBook complete with a Mosaic interface and could be viewed and manipulated by a few audience members at a time.

The demonstration also included the paper printouts of the maps. The printouts of each location, accompanied by the original map, were arranged in a set of five: full-size Iris printout of the scanned original, full-size Iris printout of the scanned microfiche, full-size Iris printout of the scanned transparency, 11" x 11" printout of the scanned transparency, and 8.5" x 11" printout from the screen on a standard desk laser printer. Questionnaires were distributed asking viewers to evaluate and compare each of the five versions.

After a brief narrative introduction to the project, the digital images were displayed and discussed by the entire audience. Afterwards the audience dispersed in small groups to look at the paper printouts and fill in questionnaires and to take a closer look at the digital files at the Pentium and the PowerBook.

The audience responded very positively to both the paper and the digital versions of the maps. The audience, familiar with Internet difficulties, was not surprised by the length of time needed to complete remote access. They were enthusiastic about the quality of resolution and found all but the largest and most complicated map (which also has some of the smallest print) fully legible on-line. They were also pleased with most of the printouts, although some noted color shifts between the originals and the printouts.

There was strong encouragement to continue work in this area, and several audience members commented that their agencies had large collections of oversized maps, architectural drawings, and other genres which would be obvious candidates for this preservation and access methodology. They were eager to move to production capability with the film-plus-digital approach.

Report prepared by Janet Gertz


(1)Scholarly Resources in Art History: Issues in Preservation. Report of the Seminar, Spring Hill,Wayzata, Minnesota, September 29-October 1, 1988.Washington, DC:Commission on Preservation and Access, 1989.

(2)Joint Task Force on Text and Image. Preserving the Illustrated Text. Washington, DC:Commission on Preservation and Access, 1992.

(3)Preservation Department. "Report on the Condition of the Collections." Chicago: University of Chicago Library, 1989.

(4)The Image Permanence Institute of the Rochester Institute of Technology is currently more than half way through a two-year project sponsored by the State of New York, "Isoperms for Color Photography," which will provide data to establish the life expectancy of color photographic film in optimum storage conditions. Cibachrome/Ilfachrome microfiche produced and stored according to national standards is already known to have at least a 100-200 year life expectancy.

(6)Anne Kenney and Stephen Chapman. Digital Resolution Requirements for Replacing Text- Based Material: Methods for Benchmarking Image Quality. Washington, DC : Commission on Preservation and Access, 1995. p.10 ff.

(7)See, for instance, the report Making Maps Easy to Read by Richard Phillips (URL http://acorn.educ.nottingham.ac.uk//ShellCent/maps/) on a project conducted at University College London at the Royal College of Art, and at the University of Nottingham.

Janet Gertz ([email protected])
(Preservation Questions)
Susan Klimley ([email protected])
(Geology Questions)

NYS Museum Bulletins | Columbia University Home Page
Columbia University Libraries/Academic Information Systems
Revised: 2/5/97 URL: http://www.columbia.edu/dlc/nysmb/reports/phase1.html