Online image search has traditionally relied on text tags and a keen eye. But new technologies can help find pictures of your boyfriend, Leonardo DiCaprio or even a perfectly worn-in pair of jeans -- regardless of the photograph's lighting or angle.
Polar Rose has developed tag-free image search for faces in photographs. Jan Erik Solem, the company's chief technology officer, says the facial-recognition technology could, for example, identify Paris Hilton whether she's in a still from a sex video or in a more demure shot from the Academy Awards ceremony.
"With an ordinary image, the computer only sees an assortment of pixels," Solem says, "so if you turn your head in a photo or change the lighting or other conditions, the computer won't recognize you."
The Malmö, Sweden-based company solves that problem with 3-D modeling technology that measures parameters such as the distance between the base of the nose and the top of the upper lip. The software also relies on statistical analyses its engineers have performed on hundreds of images to establish what head shape people with a given set of facial features are likely to have.
The program generates a 3-D head model, factoring in angle, lighting and other variations between images.
The technology should be available this spring as an extension for Internet Explorer and Firefox and as an API that developers can incorporate into photo sites.
With Polar Rose running, a tiny rose logo appears in the corners of pictures when a user loads a web page containing images. Clicking on the rose launches a search for other photos of the same person.
Privacy watchdogs worry that with Polar Rose's indexing technology, photographs intended only for friends or select groups will be as widely disseminated as Associated Press photos.
In a recent blog post, CEO Nikolaj Nyholm promised that Polar Rose will never index private photos located behind a firewall or login, and that the company's software will never be used for security or surveillance purposes. "The trick is not to turn off the technology," he wrote. "The challenge is to facilitate a way to make sure that photos that shouldn't be in our database aren't ... by restricting access or by telling us not to pick them up."
Other startups, such as Like.com and Tiltomo, have their own take on image search. Instead of faces, they have trained their software to recognize objects and consumer goods.
These programs "look at things like shape, color and texture: Is it glossy? Does it have sheen?" says Like.com CEO Munjal Shah.
The technology then converts each image into a string of numbers, its "visual signature." Users access the companies' image databases to locate similar or identical objects -- a silver Prada handbag knockoff, for instance, or an antique watch like one worn by a beloved relative.
"Gap doesn't call them 'khaki pants,' they call them 'chinos in sandstone' because they're trying to be clever," Shah says. "With this kind of search, you can find what you're looking for without having to use all those different terms."
Xerox is developing a similar "image categorizer."
Despite their varied goals, visual-search companies have similar moneymaking strategies. Polar Rose is banking on advertising-related partnerships to generate cash flow. "A text box next to an online image might tell you, 'This is a photo of J. Lo. Would you like to buy her latest CD on iTunes?'" says Mikkel Thagaard, the company's vice president of business development.
Like.com has spent the past several months forging similar partnerships, using existing product photos from retailers like Amazon.com and Zappos.com to build its database of searchable images.
As with revolutionary applications like Google Earth, however, it's hard to predict just what visual search's technological and commercial niches will be -- surprises are sure to pop up.
"We recently had an anthropologist approach us and ask if you can use this technology to find similar bones," Shah says. "As people learn more about image search, they're going to find all kinds of different uses for it that we can't even imagine right now."
