Commentary by Mark Wahl, CISA
Organizing principles for identity systems:
Metadata handling principles for press photos (20070718)
Users of digital cameras, photo editing software and photo hosting services are becoming increasingly aware of the presence of metadata in their image files. Formats such as Exif, which encodes values for technical metadata (date and time, shutter speed, color space, ...) in TIFF and JPEG, are widely supported in both consumer-grade and professional cameras and software packages.
Besides technical metadata, other categories for photo metadata are descriptive (captions, locations, persons in the photo, genre,...), administrative (GUID, date created, job id, ...) and rights (creator, credit, model releases, ...).
The International Press Telecommunications Council "Photo Metadata White Paper 2007 revision 11" includes a guiding set of four principles for metadata attached to digital images, especially photographs created by professional photographers:
- "Metadata is essential to identify and track digital images,
- Ownership metadata must never be removed,
- Metadata must be written in formats that are understood by all,
- Metadata is essential to ensure maximum image quality and image handling efficiency."
Some of the issues the IPTC noted in their white paper included
- there is a need to define mappings between different photo metadata formats with similar schema elements,
- some metadata values such as "creator" or "copyright" should be write-once (perhaps with a digital signature to detect tampering),
- file formats should include the version history of metadata, not just their most recent values,
- cameras should allow the photographer to preset some metadata so that it is automatically put onto all photos they create, and
- digital assets need globally unique identifiers.
( See also Blog posts by David Riecks from the First International Photo Metadata Conference summarize presentations by media organizations and software vendors on the importance of creating and preserving photo metadata.)
Hypothetically, a photograph such as
has a metadata record which resembles (in part)
TITLE: Destitute pea pickers in California. Mother of seven children. Age thirty-two. Nipomo, California.
AUTHOR: Lange, Dorothea, photographer.
OTHER TITLES: Migrant mother.
CREATED/PUBLISHED: 1936 Feb.
SUMMARY: Portrait shows Florence Thompson with several of her children in a photograph known as "Migrant Mother."
SUBJECT: Migrant agricultural laborers.
SUBJECT: Mothers & children.
SUBJECT: Poor persons.
SUBJECT: Migrants--California
SUBJECT: Nitrate negatives.
SUBJECT: Portrait photographs.
SUBJECT: Group portraits.
SUBJECT: United States--California--San Luis Obispo County--Nipomo.
MEDIUM: 1 negative : nitrate ; 4 x 5 in.
PART OF: Farm Security Administration - Office of War Information Photograph Collection
REPOSITORY: Library of Congress Prints and Photographs Division Washington, DC 20540
DIGITAL ID: (b&w digital file from nitrate neg.) fsa 8b29516
DIGITAL ID: (digital file from print) ppmsca 12883 http://hdl.loc.gov/loc.pnp/ppmsca.12883
There are several individual/organizational identities related to this photograph, including
- the photographer Dorothea Lange
- the organization Farm Security Administration
- the subject of the photo Florence Owens Thompson and her children
Two open questions are:
1. How should these identities be represented in photographic metadata?
- is there a controlled vocabulary of relations (e.g., "photographer", "subject", "model", "parent-or-legal-guardian-of-model", "member-of-a-group", ...)?
- should the use of unique identifiers be recommended, and if so, what forms, and how are they transferred?
- what attributes are desirable for inclusion in these identities? name? age? contact information?
- is meta-metadata needed to indicate the source, authenticity or validation of a value?
and
2. Can the four principles mentioned in the white paper have meaning for non-image (indeed non-media) data, such as identity-related information?
- Can an individual or organization attach metadata to identity-related records which they create and exchange with others?
- How do we ensure that rights information encodings are widely understood?
- A key goal of image processing workflow is to preserve metadata on photos even if the photos are manipulated/resized or converted from one file format to another - how can we prevent metadata on identity records from being lost or 'mangled' by the metadirectory/virtual directory/federation technologies which translate identity records from one format to another?