NEC announces new technology to detect pirated videos online
NEC's new technology hopes to stamp out video piracy on the Internet
NEC Japan has fired the latest salvo in war between pirates and video producers with the announcement of new identification technology capable of detecting copies of videos illegally uploaded to the Internet in a matter of seconds. According to NEC the technology achieved an average detection rate of 96 percent at a very low false alarm rate of 5ppm (5 in one million). It is also capable of detecting altered video content, such as caption overlays, camera captured copies and analog copies.
The problem of video content copyright infringement has conventionally been addressed by manual inspection. The ever-growing quantity of pirated movies and TV programs hitting the Internet makes such labor-intensive methods impractical. Proposed solutions have included digital watermarks and using image retrieval technology. However, it was prohibitively difficult to accurately inspect large databases, short content or video produced through various editing operations.
NEC’s new technology gets around such limitations by sniffing out video copies and generating a fingerprint, or video signature, of suspect content to compare to the signatures of original content. These video signatures are generated from each frame based on differences in the luminance between sets of sub-regions on a frame that are defined by a variety of locations, sizes and shapes. In this way the technology is able to accurately detect video content that was created with such editing operations as analog capturing, re-encoding and caption overlay – things that have previously been hard to detect.
The video signature files are also small in size at just 76 bytes per frame. This means that the storage space required for the matching process is small enough that a home PC (tests were carried out using a single core CPU with 3GHz clock speed) can match approximately 1,000 hours of video in one second.
The technology is also capable of detecting video scenes as short as two seconds (60 frames), which was formerly impossible when using conventional methods.
NEC Japan's new technology has been approved by the International Organization for Standardization (ISO) as the international standard for video identification as part of the MPEG-7 Video standard. Content producers will be hoping this is a case of game, set and match, in the fight against piracy, but odds are it’s just a matter of time before the pirates find a way around this latest attempt to stamp them out.
Via Akihabara News.
About the Author
Darren's love of technology started in primary school with a Nintendo Game & Watch Donkey Kong (still functioning) and a Commodore VIC 20 computer (not still functioning). In high school he upgraded to a 286 PC, and he's been following Moore's law ever since. This love of technology continued through a number of university courses and crappy jobs until 2008, when his interests found a home at Gizmag.
All articles by Darren Quick
\"The video signature files are also small in size at just 76 bytes per frame. This means that the storage space required for the matching process is small enough that a home PC (tests were carried out using a single core CPU with 3GHz clock speed) can match approximately 1,000 hours of video in one second.\"
The processing performance estimate does not look correct.
76 bytes per frame means 76 * 24 frames per sec * 60 sec in min * 60 min in hour = 6566400 bytes (~6.26 Mbytes). This means that to process 1000 hours of video per second, processing throughput should be about 6 Gbytes (48 Gbits) per sec. This kind of performance is not possible on current home hardware even equipped with a decent GPU (which, I believe, is employed). With typical movie length of 1.5 hours, size of its signature would be about 8 Mbytes, which calls for some storage.
More to that, before this analysis can take place, signatures should be created from every frame of analysed video, which would take a lot more time than the actual subsequent signature matching. This step would be necessary for each and every copy of a video online. But this, of course, could be mitigated by some preliminary analysis steps.
This is not to say bad about the algorithm itself. In fact it does look pretty feasible. And I\'m sure MPAA and the likes will be able to fetch necessary processing power. As well as establish a global universal international database of copyrighted movie materials. Membership in which might be yet another source of their corporate income, yay!
Okay, so now they identify a video more quickly than before (assuming the fingerprint exists in the database). How will this affect video piracy in any way?
I doubt this will stop video piracy. No matter what they create, there will always be some way to defeat it.
There is an easy way around that, Just encrypt the data and offer a password at the point of download or via email. Thats just one way i would think that there are heaps more ways of confusing bots..
The ONLY way to stop video piracy is the CUT THE DANG PRICES! When a movie (or computer program) in China or India costs a week\'s or even a month\'s wages - what do you think is going to happen? McDonald\'s adjusts their prices to fit the local wages. The greedy movie industry needs to do the same thing.
Over 160,000 people receive our email newsletter
See the stories that matter in your inbox every morning