From Fingerprints to Visual DNA

Computers can recognise faces in pictures, and even (it's often claimed - and disputed) filter out porn images from normal ones. But can they understand what is copyrighted content and what isn't? That's the problem facing engineers at Google-owned video site YouTube. In March, YouTube was served with a $1bn (£497m) copyright infringement lawsuit by Viacom, which claimed that copyright-protected videos had been viewed 1.5bn times on the site. In May, the Football Association Premier League launched a class action lawsuit against YouTube, and has since been joined by other content providers.The site said recently that it would roll out better technology to help detect copyright-protected video content: "Apart from being reactive and removing content when asked, what we have in place now is our digital hashing technology." When someone uploads a video, YouTube feeds the binary digits making up the file into a program that produces a short alphanumeric string representing that file. Each file's string - or hash - is unique. YouTube can compare the hash for an uploaded video against a database of hashes for copyright files. If it finds a match, it knows that someone else owns the video. View: Full story News source: Guardian Unlimited Read full story...

 


http://feeds.feedburner.com/~r/neowin-main/~3/144863947/index.php