Google Pictures Better Product Image Search

    April 28, 2008
    WebProNews Staff

Research presented by Google suggests they are on track to improving image search to a level of relevancy on par with their core web search service.

The current wave of image search services, even at Google, depends more on the surrounding text and appended metadata for pictures than on the actual images. Vaunted search algorithms might go after keywords like a terrier chasing a rubber ball, but photos make them sit back, cock their heads, and wonder when they may see a treat.

A little project referred to as VisualRank, formally known as “PageRank for Product Image Search,” promises to change this. The New York Times cited a Google presentation at the International World Wide Web Conference in Beijing, on the topic of image search and how to accomplish it.

Industry observers may be astonished to learn that Google has upper limits on its computing power, but the researchers quoted on VisualRank called analyzing and comparing all of the images Google indexed “impractical.”

They plan to focus on a subset of images and analyze them intensively. Contrary to what some may think, inanimate objects rather than people presents the big challenge to automating the recognition process for an algorithmic approach.

A breakthrough in this area of search promises a tremendous return for Google. Consumers looking for products, especially during the busy holiday season, would stand a better chance of finding what they want quickly. Couple that with Google Checkout, and Google becomes a stripped down version of online retailer