Mountain View (CA) – Google has unveiled a prototype for an Internet technology it calls “VisualRank,” an image search algorithm that it says will bring the same kind of precision its “PageRank” technology did for text-based Web searches.
In addition to changing the formula for searching specifically for image results, Google said its new technology also takes into account pictures that look similar, grouping those results together.
“We wanted to incorporate all of the stuff that is happening in computer vision and put it in a Web framework,” the New York Times quoted senior Google researcher Shumeet Baluja saying at the International World Wide Web Conference.
Google researchers say image recognition and searching has been one area of technology that has lagged behind as other areas have enhanced.
The online giant says it has the “most comprehensive image search on the Web,” but it currently relies largely on key text phrases associated with the pictures.
In structuring the new technology, the Times reports that Google first focused on specific items that ranked as top searches on Google Products (formerly known as Froogle). It then moved to matching an image search, for example, for “iPod” by scanning Google’s massive image index for pictures that matched the actual file photo of an iPod than for images with the word “iPod” attached to it.
While Google feels it is paving the future, others have been dismissive. The CEO of the online shopping search start-up Riya told the Times, “I think what they’re trying to accomplish is largely impossible. Our belief is, there [are] not large-scale solutions.”