Image quality

Blue and White Dish with a Merchant Ship, about 1510. Digital Images Courtesy of the Getty’s Open Content Program.Blue and White Dish with a Merchant Ship, about 1510. Digital Images Courtesy of the Getty’s Open Content Program.

Artstor adds content to the Digital Library through two means: managed production (the digitization of new content with production overseen by Artstor) and contributions (the addition of existing digital files generously offered by contributors).

Image Source Materials

The source materials for digital capture vary for each Artstor collection. Original source formats include 4x5 and 8x10 transparencies or negatives; black-and-white prints based on original photography made from the object; 35mm color slides, many made from reproductions in the scholarly literature; and direct, high-resolution digital photography of objects.

We are often asked about Artstor's digitizing specifications. However, it is difficult to provide "blanket" specifications that would fit all possible scenarios. Our digitizing specifications vary according to the format and quality of the source material as well as the image's content. For example, 8x10 nitrate negatives can be scanned at a much higher resolution than 35mm slides. A 4x5 transparency of graphic art with large color fields would be scanned at a lower resolution than a 4x5 transparency of mosaic details. A 35mm slide with a "sharp" image would be scanned at a higher resolution than one with a "soft" image, since scanning a soft image at too high a resolution would render the resulting digital file blurry when zoomed in upon using tools in the Artstor Digital Library.

The Image Quality-control Process

Artstor strives to offer relatively consistent image quality across collections independent of the original source media. It does so by means of an ongoing, rigorous quality-control process. Every image is viewed, tracked in a management system, and re-scanned if errors are found. Artstor staff also rigorously check contributed digital content: if the existing images can be improved or enhanced, then with approval from the contributor Artstor will often create new digital files by restoring or enhancing the contributed images.

On occasion, the original source color film scanned by Artstor may have faded or acquired a color cast. Where feasible and appropriate, Artstor attempts to restore original color to the digital scans, preferably under the guidance of the original photographer or the archive that owns the collection, and paying close attention to the distinctive physical characteristics of the photographic source. When expert guidance is unavailable, Artstor does not seek to attempt such restoration, deeming it inappropriate and the results unreliable. As a result, users will sometimes find a range of image quality, as is typically the case with other analog and digital image archives.

Continuing to Improve Image Quality

While there can be a range of image quality in Artstor, we have been proactively improving our content by working closely with museums, archives, libraries, scholar photographers, and authorized image agencies such as Scala and Erich Lessing. Image quality will only continue to improve over time as the Digital Library increases its content.

Artstor Digital Library Metadata Policy

The Artstor Digital Library aggregates collections from a wide range of sources such as museums, libraries, photo archives, scholars, photographers, artists, and artists' estates. Since each institution and individual can have differing uses of and requirements for metadata, the cataloging and descriptive data we receive vary greatly in the use (or absence) of standardized vocabularies, in the choice of terminology, and in the metadata schema used to organize the data. Even within a single institution, the focus, and thus the cataloging, may vary. A museum conservation department's terminology and focus will differ from that used by the curatorial departments, and a curatorial department that is responsible for archaeological material will collect and store information in ways that differ significantly from, say, a modern painting department. Since the holdings of museums, artists, archives, and other contributors are most often unique, those cataloging methods and controlled vocabularies have often been created for the local environment, and not necessarily to enable sharing those collections with a broad range of educational and scholarly users. While this outlook is changing as the usefulness of sharing descriptive data and using shared standards is recognized, Artstor still receives hundreds of thousands of records from contributors whose descriptive data were created well before the potential value of sharing was recognized and before any generally accepted standards were in use; to this day, the community does not share any one standard and data continues to be heterogeneous. The variation in quality, authority, and consistency that one sees within the Artstor Digital Library is the result, then, of our bringing together contributions from a wide variety of sources.

Artstor, by necessity, considers the data from multiple perspectives. From one vantage point, the data reflects the interests, concerns, and point of view of the contributing institution. In this sense the data serves to describe the works and images in terms selected by the original source. Artstor makes some attempt to preserve, almost in an archival sense, the characteristics of the source data. Yet this attempt must often be modified to bring all data sources into enough harmony that search and discovery across the Digital Library are as effective as possible. In addition, our users expect the descriptive data to be accurate and "correct." Satisfying this view brings with it special challenges in areas where facts are scarce and/or scholarly opinions differ. As information for images, objects, architecture, and other materials changes over time, and as scholarship continues to evolve, to uncover new findings, to spark new debates, or to trigger reassessments of attributions, dates, and other descriptive information, Artstor works with our numerous contributors to update and refresh the metadata records in the Digital Library. We are committed to facilitating the ongoing work of educators and students as well as scholars and researchers (who welcome and often benefit from varying metadata records that might reflect different modes of thinking, historical trends, or phases of scientific research). For these reasons, we attempt to preserve and share as many versions of images and metadata records as are provided to us by our international contributors, while foregrounding the highest resolution image (and its associated metadata record) available. The descriptive data records one sees in Artstor are thus a compromise; the measures described below are intended to enhance discovery and access for all of our educational users—scholars, curators, educators, librarians, and students.

Current and Ongoing Metadata Work

Metadata Enhancements

Artstor enables all collections in the Digital Library to be searched and browsed by object-type classification (e.g. painting, architecture, etc.), country/region, and earliest and latest date. This is accomplished by adding information to records rather than altering the source data. The classification terms are applied from an in-house, controlled list (painting, sculpture, etc.); the country terms are from the Getty Research Institute's Thesaurus of Geographic Names (TGN); and numeric earliest and latest dates are created for each record. Our Artstor Advanced Search and Browse functions depend on these uniform access points to improve the user's ability to search and browse across all the Artstor collections. These enhancements are not visible within the metadata records but rather operate "behind-the-scenes" within the database to facilitate searching and browsing.

Clustering Images

With over 1.5 million assets (and growing), it becomes increasingly difficult to readily find all the images of the same work. We often receive duplicates or details of the same work from an individual contributor or from multiple contributors. To help alleviate this problem, Artstor is grouping or "clustering" duplicates and details representing a unique work. We cluster these images and details of the same work "behind" the highest quality image of the whole work we have available.

Associated Images

Drawing upon data from user-curated image groups, Artstor is also able to present researchers with the option of discovering "associated images." Through mathematical analysis, Artstor can determine which images have been "saved" in association with other specific images in groups that have been created by instructors. We assume that images saved repeatedly in such "associated" groups are related in ways that are useful to teachers and scholars. These collaboratively filtered groups conveniently bring together many works associated with the lead images. At times, these juxtapositions in the "long tail" can be surprising, original, and intellectually stimulating.

Controlled Vocabularies

Because of the diversity among Artstor collections, standard vocabularies or thesauri are particularly significant for discovery. Artstor uses the Getty Research Institute's Union List of Artist Names (ULAN) to match artists' names with an authoritative creator record. Through these matches, links are established between the source creator name and the ULAN creator record which allows, for example, a user searching for works by Gerrit von Honthorst to find images that have the name Gherardo della Notte or Gherardo Fiammingo in the source data record. We intend to extend this matching of source data to external vocabularies into other areas of information such as repository names, geographical locations, styles, and periods.

Future Plans

Contributing to ULAN

In addition to matching our data records to ULAN, Artstor is adding authorized creator names to an authority we call the Artstor Name Authority (ANA). These new names, as well as those that are added to ANA through JSTOR Forum, will be submitted to the Getty Research Institute to add to ULAN. Since libraries and institutions around the world rely upon the Getty vocabularies, contributing new creator name records will benefit the communities which utilize the Getty's name authority for cataloging cultural heritage objects and images.

Work Authority File

Artstor's JSTOR Forum data model includes a work authority "section" that is partly based on the structure of the Getty Research Institute's proposed Cultural Object Name Authority (CONA). Artstor and the participants in JSTOR Forum will develop a shared work authority file that will be made broadly available and provide material for CONA. Work authorities and the more complex data structure supporting the work concept will enhance cataloging and perhaps even allow images to be displayed in a more structured way in the Artstor display and discovery environment. This more complex data structure might also facilitate searching or filtering by more refined categories than is currently possible with our flat metadata schema.

Expert Tagging

Artstor, as a community-built resource, is working on ways to facilitate the contribution of useful information by our community of scholars, visual resource professionals, and other knowledgeable users. Such a tagging feature would allow expert users to contribute cataloging information, comment upon existing information, and provide scholarly commentary about the images, underlying works, and/or the data provided in the current record. There are a number of challenges to the use of tagging and expert commentary, not least of which is determining a good way to mediate and present alternative information and varying interpretations. Still, we recognize that it is essential, given the ever-expanding scale and scope of the Artstor Digital Library, for the descriptive data to improve and expand in as many ways as possible.