Google researchers on Monday unveiled a bold new image search technology. During a presentation at the International World Wide Web Conference in China, the Google scientists described the company's
new VisualRank technology as blending image-recognition software techniques with criteria for weighing and ranking images.
The scientists noted that despite years of attempts, image search
remains an unsolved mystery-even though the likes of Google and Yahoo offer image search, the results are largely based on how the images are tagged, rather than on how they look. While progress has
been made on face detection, finding objects has proven to be far more difficult.
"We wanted to incorporate all of the stuff that is happening in computer vision and put it in a Web
framework," said Google researcher Shumeet Baluja, who made the presentation with fellow researcher Yushi Jing. Their paper, "PageRank for Product Image Search," focuses on the 2000 most searched for
products on Google Image Search, and creates a scoring system for image relevance. It found that that the new algorithm returned 83% fewer irrelevant images. The researchers claim that Google Image
Search is now the "most comprehensive image search on the Web."
Read the whole story at The New York Times »