Similar applications of CBIR technology in digital
libraries include the University of California-Berkeley's
Digital Library Project (http://bnhm.berkeley.edu), the
National STEM Digital Library (ongoing), and Virginia
Tech's anthropology digital library, ETANA (ongoing).
While these feature-based approaches have been
explored over the years, an emerging new research
direction in CBIR is automatic concept recognition and
annotation. Ideally, automatic concept recognition and
annotation can discover the concepts that an image con-veys and assign a set of metadata to it, thus allowing
image search through the use of text. A trusted automatic
concept recognition and annotation system can be a good
solution for large data sets. However, the semantic gap
between computer processors and human brains remains
the major challenge in the development of a robust automatic
concept recognition and annotation system.
A recent example of efforts in this field is
Li and Wang's ALIPR (Automatic Linguistic Indexing
of Pictures—Real Time, http://alipr.com).