Google has revealed details of new image recognition software methods which will be used to facilitate online image searching.Speaking at the International World Wide Web Conference in China, Google scientists described VisualRank, an algorithm for combining image recognition techniques with methods for ranking images that look the most similar, the New York Times reports.
According to the Google experts, image analysis is still a largely unsolved problem in the field of computer science.
As looking at all indexed images would be impractical, Google has reportedly focused on a subset of the images it has catalogued in order to analyse and compare the digital images.
However, the scientists stated that its Google Image Search is the "most comprehensive image search on the web".
Google says it has concentrated on the most popular product queries on its product search, including words such as Xbox, Zune and iPod.
A team of 150 Google employees has been working on creating a scoring system for the relevance of images related to these products.
According to the researchers, as a result of the work they have done, 83 per cent fewer irrelevant images were returned in searches.
Commenting on the exercise, Shumeet Baluja, a senior staff researcher at Google, said: "We wanted to incorporate all of the stuff that is happening in computer vision and put it in a web framework."
Originally announced in 2001, Google Image Search allows users to search the internet for image content.
