In one of the readings for this week, an article “Towards an Application Profile for Images,” Mike Eadie describes the IAP (Images Application Profile) project funded by the Joint Information Systems Committee. While the article dates from 2008, and it is unclear without further investigation if this application profile was ever adopted, the article brings to light some of the challenges of creating image metadata.
What is an Application Profile?
An application profile is a metadata vocabulary built from preexisting schema and adapted for a specific application. In the IAP project described in the article, an attempt was made to develop an application profile based on Dublin Core for image retrieval across institutional repositories among different academic disciplines.
Creating Metadata for Images
Metadata description of images can be a challenge because it usually requires manual human input of a description of the image. As noted in other readings, human input can be time-consuming and possibly introduce errors in the metadata. There is content-based image retrieval (CBIR) whereby machines “read” an image for data collection. While this software has come a long way since the time of this article in 2008, CBIR is imperfect and incomplete in image description. Eadie describes the semantic gap—the distance between description by machine and what humans can provide, and need, for image description.
Images require several considerations not found in text description. According to Eadie, digital images require at least three levels of description:
- Technical information relating to image
Content depicted in image
Abstract meaning and interpretation
Another aspect unique to images is that they can have relationships with other objects and images. Images can appear in other objects such as books or presentations.
Images come from a wide array of academic domains and may require different terminology for user access—for example, the difference in vocabulary and search terms an art historian would use as compared to a biologist looking for images in an academic repository.
The article goes on to describe the scope and methodology of the IAP project. I think the points made regarding differences in image metadata versus text metadata were interesting to consider. I wonder what happened to the Image Application Profile. Is anyone familiar with this?