Establishing Spatial Resolution Requirements for Digitizing Transmissive Content: A Use Case Approach

Guidelines for the digital conversion of transmissive content (negatives and positive slides) in the cultural heritage community have lagged behind those for print content. The primary reasons for this are twofold. Unlike print material, transmissive content is generally an intermediate format (as with negatives) or requires a viewing mechanism (such as a projector). In either case, there is no standard for the viewing of the object. The second challenge for digitization of transmissive content is, in large part, a result of the ambiguity of the visual output for slides or negatives. Typical guidelines for the resolution in digitizing transmissive content have concentrated either on the limits inherent in scanning equipment or attempted to base resolution on the microstructure (e.g., grain) of the negatives or slides, and not on the actual image information content in the film. With special regard to the latter and in response to a project to digitize negatives from the Farm Security Administration (FSA) collection at the Library of Congress, the first part of this study investigated methods for establishing scanning resolution requirements for B&W silver-gelatin film negatives from the early to mid 20 Century based on defined use cases. The use case described by the curators of the FSA collection for resolution was based on the information content of the original photographic image and not on a film’s granular microstructure. In other words, actual image information captured and not the inherent grain structure itself. This study describes the methodology used for determining the limiting resolution of film capture, and presents the results of the study for the FSA collection, as well as the application to all similar transmissive content. Examples will be provided to demonstrate the utility of this approach. In addition to silver-gelatin B&W film negatives, the second category of transmissive photographic originals studied was early color positive processes, specifically Autochrome and Dufaycolor. The nature and perception of the image for these processes is very different than the typical image and silver grain structures of typical B&W negatives. In addition to the color information, these photographs represent a very distinct class of original and different use case requirements. Investigation of these materials provided an opportunity to expand the research methodology used for determining appropriate scanning resolution. Introduction One person’s flower can be another one’s weed. It all depends on the intended use. This same logic applies to determining required levels of spatial detail when digitizing image content from any form of photographic original, be it reflective or transmissive. To turn a phrase, one person’s signal can be another one’s noise. Examples of this for reflective media are simple and few. An obvious one is choosing to digitize for paper structure (i.e. tooth) or simply the content provided by the original marking process. The latter is less demanding and almost in all cases more manageable and economical. Halftone structure is yet another. Are these flowers or weeds? Both can be considered of informational value, depending on who is asked, or an annoyance. The division lies in what can be considered of image value versus artifactual value. Examples of transmissive media offer similar but more complex challenges. We have chosen to concentrate on these in this paper for several reasons. They are: Lack of sound scientific data on the image information content of archived silver-halide film content Unusual spatial detail characteristics, including microstructure, of past transmissive photographic processes ( e.g. Autochrome, Dufaycolor) An apparent and unsupported trend to choose unusually high sampling frequencies (2500-4000 ppi) as a standard for archival film digitization In addition to resolution limits, the structure of slides or negatives may result in a viewing experience that is unique to the class of material. In such cases, the requirements for digitization may require the ability to represent the unique aspects of the analog content, its artifactual value. We have taken two historic formats to explore this use case Autochrome and Dufaycolor. Both formats present a unique viewing experience resulting from the microstructure based on the manufacturing process for each format. The objective of this study is to demonstrate methodologies, preferably analytical ones, for establishing and verifying digital imaging requirements for scanning resolution of photographic negatives and slides. As described above, two use cases for digitizing slides and negatives will be studied. They areA Image information The digital image is required to capture the full image information content of the original scene or object captured in the original photographic image. In this use case, increased resolution requirements for the digital image will not yield any additional detail about the scene or object that was photographed. There is a belief that large format B&W negatives require high sampling frequencies because they have such a wealth of information in them. This information however is actually spread across a very large format with concomitant low image information per unit area on the negative. It is this packing density that determines required sampling frequencies (i.e., ppi). There is anecdotal conjecture by optical engineers that while the film itself was capable of high resolution, the lenses in the cameras used through the first half of the 20 century were not really that good and acted to limit the effective resolution of film images from those times. B Artifactual information The digital image is required to capture the qualities of the negative or slide that is responsible for the characteristic qualities of the particular medium. An example of this is the Autochrome color process, where the size and distribution of the dye granules result in a photographic image with a visual quality similar to a pointillist painting. The requirements for this use case are independent of the requirements for capturing scene detail as in the use case described above. Experimental Approach There are four primary components to this study. They are: Media selection Measurement device calibration Data collection Data analysis and discussion These items are serial in nature. Details of each follow. Media Selection To limit the scope yet create the most benefit, only first generation transmissive media intended for pictorial purposes was considered. While no color negative materials or microfilm were included, the analytical methodologies outlined in this paper for determining resolution levels can also be used for these materials. The largest portion of the materials considered was B&W silverhalide negatives created throughout the 19th century. The authors attempted to sample from a wide variety of formats. These included 35 mm through 8x10 formats. Exploratory research was also done on media from two nontraditional transmissive media types, specifically Autochrome and Dufaycolor imaging processes. These were chosen because of their unusual microstructure and the way this structure can be discriminated from true image content information. Measurement Device Calibration All measurements of image detail and microstructure of the photographic originals were made with a Meiji MT8000 microscope with 20x and 50x plan objectives, equipped with a ProgRes C5 5.0 Megapixel CCD Camera used to capture illustrative images and for indirect measurements. Direct visual measurements were conducted using Mitutoyo Digimatic Series 164 Micrometer Heads, having an accuracy of 0.00015′′ for XY stage movements. The calibration of the ProgRes camera for all objectives was done using variable frequency glass Ronchi Rulings (5 lp/mm to 200 lp/mm). Calibration measurements were taken across a series of line pairs using the camera live view, taking four readings for each objective. The measurements were then verified by taking a direct measurement with the stage micrometer to prevent any gross errors that could be made by mistaking the ruling frequency. Calibration of the microscope/camera was performed monthly. The objectives were also calibrated immediately prior to making the measurements used in this study. In order to establish resolution requirements, a calibrated device is required to accurately measure resolution. To demonstrate the viability of our approach, we chose two commercial off-the-shelf (COTS) transmissive scan devices. One was a Kodak iQsmart3 scanner at the U.S. National Archives at College Park. The other was a BetterLight camera/scanner at UC Berkeley. Both were calibrated using a film target that had been previously established suitable for accuracy to 10000 ppi and is often used for validating microfilm scanner performance. It is pictured in Fig. 1. The true optical resolution of the cited scanners was determined through SFR analysis as described in ISO 16067-2 using slanted edge Spatial Frequency Response (SFR) protocols. The Kodak iQsmart scanner was generally used in the 3500-4000 ppi range. Measurements confirmed that it provided true optical resolution at whatever sampling frequency was selected, up to 5000 ppi. This level of resolution is more than sufficient for capturing the information and microstructure of historic images on transmissive media. The authors realize that high-end devices like the Kodak iQsmart3 scanner are often unavailable to most users (this scanner and similar models have been discontinued). To that end we also used a BetterLight scan back to demonstrate the use of typically available equipment in performing the characterization techniques described in this paper. Unlike the flatbed technology used in the Kodak iQsmart, the BetterLight was mounted on a copystand with a backlight illuminator. This parallel portion of the study was done at UC Berkeley’s Moffitt Library by a staff photographer. Using a 50 mm lens, a true 4464 ppi was achievable. Several lenses were tried in order to achieve this level of performance Example SFRs of the scanner’s used in this study are shown in Fig. 2. SFRs from both scanners were very well behaved and reflected the true optical resolution of the selected sampling frequency. These SFRs were then used as scanner correction factors to determine the spatial information content in any selected negative. The way the data was selected from the negatives to determine this is described next. Figure 2. SFRs for scanners used in this study Figure 1. Transmission Scanner Resolution Calibration 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1