Google has developed new search technology intended to do for digital images what PageRank software did for searches of webpages.
When the PageRank algorithm was developed nearly a decade ago it revolutionised the way web searches were carried out. Now the company is hoping to repeat its success for online image searches.
PageRank for Image Search was unveiled by two Google scientists at the International World Wide Web Conference in Beijing last week.
Although presented as a research paper, Google hopes that the new algorithm will overhaul the text-based methods currently used by Google's image searching technologies.
"Our experiment results show significant improvement, in terms of user satisfaction and relevancy, in comparison to the most recent Google Image Search results," the scientists wrote in their paper.
Image searches often result in unsatisfactory returns, selecting results based on the corresponding text round the image rather than the content of the image itself.
Google's new model, also known as "VisualRank", will endeavor to actually understand the image.
The two Google scientists tested their algorithm using images retrieved by Google's 2,000 most popular product searches, including words like iPod, Xbox and Zune.
It then compared the top 10 images with the standard Google Image Search results. The researchers said the retrieval returned 83 per cent less irrelevant images.
While it is still in the research stage for Google other companies are using image recognition software.
Riya, based in Silicon Valley, California, is now offering an image-recognition shopping engine, known as Like.com, that locates products on sale across the web.
http://images.google.com www.like.com 
Topic: Google aims to improve image search (Read 793 times)