• E-mail this page
• |  Print this page
• | 
• | 
• | 

Google Refines Its Image Search Technology

Having won the text search war, Google may soon take the lead in image searching. In a paper published last week, Google researchers describe how they were able to use image recognition technology to reduce the number of irrelevant images returned through a Google Image search by 83%.

For Google, improved search relevancy helps improve click-through rates, and that means more traffic and revenue for Web sites and for Google. "When examined over the set of 130 product queries, the images selected by our approach to be in the top 20 would have received approximately 17.5% more clicks than those in the default ranking," the paper explains.

More clicks mean more searchers who have found what they're looking for, and that's only going to extend Google's search lead. Consider that Internet metrics firm Hitwise on Monday reported that in the U.K., the top 100 Google Internet properties were responsible for more than one third of all Internet traffic to Web sites during the month of March.

That's the kind of market power that drives companies into vertical markets.

Google isn't the only company working on image recognition technology. There are startups like Riya and Pixsta trying to improve image searching. But they don't have as many users as Google, so their breakthroughs need to be much better than what Google comes up with, or different enough to justify co-existence, if they're to last. Chances are the startups that thrive will tackle image recognition problems specific to industry verticals, like medical imaging.

Xerox pursues this strategy. At a press tour of its Palo Alto Research Center (PARC) on Tuesday, Christopher Dance, a laboratory manager at Xerox Research Centre Europe (XCRE), demonstrated the company's hybrid categorization technology for helping to identify images. The technology is hybrid in the sense that it combines image recognition algorithms with semantic analysis of text associated with images. In so doing, Dance claimed the system outperforms image classification schemes based solely on either text or image data.

An Italian journalist attending the demonstration asked Dance whether Xerox might make its technology available to the public, as Google has done with its services. Dance said he didn't think so, because doing so would undermine Xerox's ability to sell its technology as a service to specific business markets. "If we released it to everybody, it wouldn't have the same value," he said.

If you're Google, you can give technology away to everyone and still end up with a lot of value.

How much is it worth to establish the expectation that technology should be free, making companies with traditional business models seem somehow greedy for refusing to share?

« ReiserFS Without Hans Reiser, Continued | Main | Pay Isn't The Only Gap Between Male And Female Tech Pros »