Google is in the process of developing new search technology to enable it to recognise and rank images more accurately.
The search engine hopes to make image searches as accurate and relevant as text search by using technology to read the image.
Currently, finding pictures through Google Image Search is currently imperfect, as the search engine has to rely on the text associated with the image, which may often bear no relation to the picture.
Now, Google researchers have devised a new algorithm dubbed Visual Rank that will looks for visual themes across a range of photos, ranking images according to how similar they are to other images that contain the same theme.
This new technology is still at the research stage, though early results are promising – Google tested their new algorithm using images associated with the search engine’s 2,000 most common product searches.
The researchers then compared the top 10 images with the standard Google Image Search results, claiming that the new image search returned 83% less irrelevant results than the current system.
Google is not the first company to develop new image search technology though. Silicon Valley firm Riya developed technology for e-commerce site Like.com that allows users to search for fashion items from related pictures.
In addition, Swedish startup Polar Rose has been working on photo search and facial recognition technology for use on the web.