Riya, a fledgling site for organizing and searching photos that has enjoyed early popularity, is developing a search engine that will let users find digital photos across the Web by recognizing similarities in images.

(click image for larger view) A face like no other

Riya 2.0, scheduled for release in the fall, will use a technique that relies on hundreds of algorithms to look inside a photo and break it into grids to identify objects, Riya CEO Munjal Shah says. The software examines each face or object and turns it into a math equation that has about 2,000 parameters, Shah says. Within those parameters, the software can identify specific characteristics about noses, chins, and foreheads, for example, by looking at the amount of space between a person's hairline and eyebrows. The software also can recognize text within images, such as the text on a sign posted on a building in a photo, according to Riya.

Since the site launched in March, Riya users have uploaded more than 7 million photos to its servers, where they have remotely trained the system to recognize images and tag them. A user with hundreds of family photos can categorize them by members. The search engine will expand this ability beyond the Riya site. "Many people really wanted the ability to search the Web for all the pictures of them or a person that looks similar to another person," Shah said.

The technology may not be limited to consumer fun. Riya's image search engine could appeal to marketing and advertising companies, Susquehanna Financial Group equity analyst Malindi Davies says. "If you have a more efficient way to tag or recognize objects in photographs that's less human-capital intensive, that's a great technology for stock-image companies like Getty Images, Jupiterimages, or Corbis Photos," Davies says. Riya received $15 million in a second round of venture capital funding from Bay Partners earlier this year.

But while Riya's approach is unique, it faces competition from established photo-search sites, including Yahoo's Flickr and Google's Image Search.