No Subject

["Hamaker, Chuck" <cahamake@email.uncc.edu>]

April 03, 2001 issue
Business 2.0

Image Problem
by David Orenstein

Finding the right picture on the Web has been a 50 billion-to-1 shot. The
odds are finally starting to come down.  
http://www.business2.com/content/channels/technology/2001/03/26/28468

The digital imaging industry has developed a crucial new standard that
will allow computers to efficiently find pictures based on metadata, or
text that describes each image. And researchers are working to make
entering that text as effortless as possible. Meanwhile, the industry is
struggling to find a sexier solution, using pattern-recognition software
that can "look" at images and then find others that resemble them.
Although neither method functions smoothly yet-indeed they both contain
serious weaknesses-businesses are already probing the potential of the new
technologies:

Metadata is nothing new in computing or even in photography, but it's
especially easy to use when available in a standard format. Therefore,
change came when the Digital Imaging Group
<http://www.digitalimaging.org>, a consortium that includes Canon ,
Eastman , and others, created the DIG35 standard, a standard simple and
universal enough to succeed where proprietary metadata systems and EXIF,
the current industry standard, failed.

To spur adoption of DIG35, companies began rolling out software last fall.
At Comdex/Fall in Las Vegas, Kodak reported it will give away its once
proprietary Picture Metadata Toolkit software. The software provides a
standard way to access image metadata in the DIG35 and EXIF formats.
Meanwhile, German image-management software company Canto Software
unveiled a DIG35 filter for its Cumulus software in January at Macworld in
San Francisco.

Metadata automata 

Because many photographers (especially amateurs) refuse to type in
extensive metadata, imaging companies aim to make it easier to do. Digital
cameras that use EXIF already embed information such as the time, date,
lens type, and aperture settings of a picture, says Tony Henning, a senior
analyst for Future Image. A few professional models with global
positioning system chips even store the exact location where a photo is
taken. The real frontier, however, is finding a way for more sophisticated
data, such as the actual content and significance of a picture, to be
entered automatically.

Convera is devising ways to automatically identify metadata within photos
and videos, says Dan Agan, a senior vice president. One path being
examined: Teach a computer to recognize descriptive text such as "Welcome
to Grand Canyon National Park." Another tactic: Convert spoken words into
text.  Major media companies such as ABC News Internet Ventures and
Discovery Communications have been using a combination of proprietary
metadata and image-recognition software to cut the costs of managing their
huge image databases. By midyear, Lycos will bring that technology to the
Web. The site plans to add image-recognition software to its
picture-finding engine, which previously could only search the captions of
pictures in its vast picture galleries.  For rest of article see link at
top.

---end---