Where Oh Where Art Thou, Digital Image?

December 05, 2007

Photoalbum I’m one of those. I immediately upload my digital images to a photo-sharing website and send the link to all of my friends and family. To date, I have 53 albums and probably thousands of images going back to 2004.

The sharing part is great. But the archiving and then finding part is a royal pain in the arse. It’s difficult to search the stack and locate the exact image, unless you rename each individual image. And frankly I, and lots of other people out there, don’t have time for that. Wouldn’t it be great if your camera and computer would work together to automatically handle some of that tedious archiving?

Sure thing, said Jiebo Luo, a senior principal scientist with the Kodak Research Laboratories in Rochester, NY.

“I have the same frustration,” he said. “I try to organize photos by folders. Some people name each picture, but that’s too much work.”

Luo and his team at Kodak are just one group of researchers (there are others at Yahoo!, Google, and Microsoft doing the same) trying to solve this problem.

I called Luo to ask him why annotating, archiving, and searching of digital images is still largely a manual task akin to placing prints in a photo albums.

The problem, said Luo, is semantics. We can do two important things: recognize objects like sky, water, and sand, and then infer that those objects combined equal a beach. It’s just plain difficult to translate semantic concepts into computer language, which is based on numbers.

Researchers are working on a solution though and most of them are trying to combine contextual information that can be gathered automatically by the camera with more advanced computer algorithms that can squeeze hints out of numbers.

For example, lots of GPS-enabled mobile phones have cameras these days. A group at Yahoo! Research Berkeley has developed Zone Tag, which uses GPS location data to automatically tag, or suggests tags, for images being uploaded to Flickr. Anyone can download the program and try it out.

Conventional digital cameras also have setting information such as exposure, shutter speed, subject distance, and scene mode (portrait, landscape, night, sports) that could be saved in the digital code for each image. Combined with GPS and time information, a computer could narrow in on the fact that a certain group of images were taken at Lambeau Field in Green Bay under the “sports” setting and, that with all of the green and gold pixels, that this is most likely a Packers game.

If you took your pictures with a digital camera, you can try Penn State’s program, Automatic Linguistic Indexing of Pictures, which automatically tags images as they are uploaded. I wrote a piece about ALIPR on Discovery last year. But since then, the system has been improved with technology called Tagging Over Time, a software system that improves the tags by learning from user interactions.

Although Zone Tag and ALIPR are still research projects, some of the technological advances that will make auto-tagging easier are making their way to the commercial market, say Luo. Take face recognition. Nikon and Cannon both have face recognition software that finds the face of your subject and uses it for auto focus. The next step, said Luo, would be to find the face and then recognize that it’s Aunt Bea and label the image accordingly.

I guess the message is: Hang in there. Help is on the way.




Tracy Staedter pulls the levers and pushes the buttons behind the curtain of the Discovery Tech Web site.
discovery channel tech





Advertisement

SITE SEARCH
SUBSCRIBE TO OUR NEWSLETTERS
CREDITS DCL |
DISCOVERY SITES Discovery Channel / TLC / Animal Planet / Discovery Health / Science Channel / Planet Green / Discovery Kids / Military Channel /
Investigation Discovery / HD Theater / Turbo / FitTV / HowStuffWorks / TreeHugger / Petfinder / PetVideo / Discovery Education
VIDEO Discovery Channel Video Player
SHOP Toys / Games / Telescopes / DVD Sets / Planet Earth DVD Sets / Gift Ideas
CUSTOMER SERVICE Viewer Relations / Free Newsletters / RSS / Sitemap
CORPORATE Discovery Communications, Inc / Advertising / Careers @ Discovery / Privacy Policy / Visitor Agreement
ATTENTION! We recently updated our privacy policy. The changes are effective as of Tuesday, October 30, 2007. To see the new policy, click here. Questions? See the policy for the contact information.