Can Google’s technology automatically analyze what’s inside an image? I mean is it really possible to automatically analyze an image subject then rank it according to the derived description without human intervention? Well for Google’s two scientists who presented a research paper in China last week, it might be possible.
During the International WWW Conference in Beijing, two Google Scientists presented their research paper on “Page Rank for Product Image Search.” The research paper will apparently remove Google’s dependence on the alt text that is associated to an image describing its subject and content. Through VisualRank, an algorithm that blends image-recognition software methods with techniques for weighting and ranking images that look most similar, Google’s new image search will rank images automatically based on the derived analysis of their subjects and contents.
If the results of the research paper pushed through, we might see the time when humans will no longer be needed to describe images through metadata. Computer will do the content analysis of the image which will then be crawled by Google’s image searching algorithms and applies the same PageRanking technology it used for web pages to rank these images.
Arrington got the full PDF file of the report. Check it out if you want to read further.