Google say they have developed a image search software intended to “see” the content of images and do for digital images on the Web what the Google’s PageRank software did for searches of Web pages.

On Thursday at the International World Wide Web Conference in Beijing, two Google scientists presented VisualRank, an algorithm for blending image-recognition software methods with techniques for weighting and ranking images that look most similar.

They conducted a series of experiments based on the task of retrieving images for 2000 of the most popular products queries. Their experimental results show significant improvement, in terms of user satisfaction and relevancy, in comparison to the most recent Google Image Search results.

Image search at the major search engines today largely relies on looking at words that are used around images — on the pages that host them, in image file names and in ALT text associated with them. No real image recognition is done by any of the majors. Search for “apples,” and they haven’t actually somehow scanned the images itself to “see” if they contain pictures of apples.

The method in Google’s paper changes that. In short, a group of images retrieved for a query using traditional search methods is then further analyzed. Image recognition software finds which images in the group seem most similar to each other. It then estimates “visual hyperlinks” between them to produce a final ranking.

The last part is important. No actual hyperlinks on the web are used to rank the images, if I understand the paper correctly, other than in the first traditional retrieval process. Instead, the algorithm guesses at how the images would be linked together, with those being most similar having more virtual links to each other. As a result, the most “linked to” images are calculated to rank first.

Get the full story at searchengineland.com