"Visual Google"

This was developed by Hartmut Neven, a researcher here at USC's ISI lab in Marina Del Rey - we should propose some content and user studies??:

Neven has applied for a patent to cover the use of image-recognition software on mobile phones and has started cutting deals with various companies. Vodafone Japan and NTT DoCoMo offer wireless video-messaging services powered by Neven Vision technology. Vodafone's MovieMask, launched in July, recognizes your changing expressions as you look into the camera and adds the appropriate special effects, like tears or sparkles. DoCoMo introduced a similar service called Face Stamp in November. Neven Vision expects at least three European cell phone carriers to make the technology a standard feature next year.

Meanwhile, the company is developing a security application that would use biometrics - facial features, skin texture, and iris pattern - to authenticate purchases made via cell phone.

And this fall, after two years of development, the company is rolling out its most ambitious service, what Neven describes as a "visual Google." The company has tweaked its facial analysis algorithms to identify anything from a Coke can to the Mona Lisa, barcodes to kanji. By linking this object-recognition software to a database of images, Neven aims to build a search platform for phonecam users. Don't know what something is? Snap a pic and the service sends back a match within 10 seconds.

The technology will debut next year in ads that offer, say, $1 million to the millionth person to submit an image of a can of Coke. Travel guides are next: Snap a picture of the Pantheon to learn its history, or click a road sign you need translated from German to English. "The system hyperlinks the visual world," Neven says. "Eventually every building and object will be in the database."

Wired 12.12: START