Image service project

Participants and purpose

The CHS Technical Working Group is working with participants in the Multitext of Homer project to develop an integrated Image Service.

The Image Service brings together queryable metadata about a set of images (such as those being collected for the Venetus A MS of Homer) with a service for manipulating images.

This combination allows applications to identify, retrieve and perform simple manipulations of images.

Technologies

Metadata about images will be exposed to querying through the Collections Service API. A separate Image Manipulation service initially allows retrieval, scaling ("zoom in/zoom out"), and selecting sections of images ("panning").

Image manipulation is implemented with the ImageJ library developed by NIH. (See ImageJ home page.) In contrast to the service developed for the Fedora CMS, images are identified and referred to by a stable, canonical identifier (rather than by URL). (See documentation for the Fedora image manipulation servlet.)

Current status

Neel Smith is currently adding a relational database backend to convert canonical references to pointers to physical images (local files, or URLs).

Questions/TBDs

Beyond retrieval, scaling and selecting, are there other manipulations we should plan for the first version of the Image Manipulation service?
What metatata about binary image objects should be accesible to the Image Manipulation service (as opposed to artibitrarily rich metadata in an associated Collections service)? File format? Color model? Should this data be integrated into the image's binary data (e.g., in EXIF format)?
The Collections service with image metadata and the Image Manipulation service must use a coordinated set of identifiers. Like other Collections, these should have two parts: a registry authority, and the identifier string; in combination, the two are guaranteed to be unique. We need a Regsitry service at CHS for registering Collections naming authorities.