The test will begin next month with hopes that the software, designed to recognize copyright content in videos, will be ready to roll out later this year, the company said.
Video recognition is more complicated than audio fingerprinting, Maxcy said. It involves extensive indexing of images from videos and collecting the images in a database. It also involves compiling the rules associated with each piece of content as set by the copyright holder.
For instance, a TV network could tell YouTube to pay a royalty whenever one clip is uploaded to its site but block another that is unauthorized.