The advent of Internet-based video is upending the business model that Hollywood and TV program producers have embraced for decades. Although the opportunities have yet to fully flesh themselves out, web-based threats are becoming all too clear and are foremost on the minds of most studio executives.
Content piracy and unauthorized distribution of movies, TV shows and big ticket events seem to be everywhere; the recent high-profile pre-theatrical release of Michael Moore’s “Sicko” on YouTube has only escalated efforts by content providers to find the means of searching the web for copyrighted video material.
But, unlike text, and even still images, video is fluid and entails multiple frames per second, making the task of identifying specific pieces of copyrighted content difficult, if not impossible, from simple searches alone. Washington, DC-based LTU Technologies, however, has a plan to help content providers scan the web for proprietary video based on its “image DNA” technology.
LTU has developed this algorithmic system after years of working primarily with law enforcement officials such as the FBI, the US Immigration and Customs Enforcement Bureau and the French Police to locate child pornography or terrorism-related images or find stolen art. This image DNA technology is an “exclusive descriptor that indexes, recognizes, and describes images by their visual content, instead of solely by keyword or file name,” giving LTU a leg-up over other kinds of recognition systems, such as digital “fingerprinting,” the company contends.
With digital fingerprinting, images or video frames are merely assigned unique identifiers so that when pirated versions of content appear on the web, the content can be identified only by its unique number. Image DNA, on the other hand, tracks “what’s going on in the totality of the clip,” Kevin Smith, LTU’s Vice President of Sales for North America said.
It’s superior to other methods of locating content because it can find content that has been altered, so-called “cloned” content, where, for example, background colors have been shifted, or content that hasn’t been fingerprinted or otherwise watermarked, or content where the watermarking descriptors have been corrupted, according to Smith. That’s because the video images themselves are encoded in a form of 3-D mapping that allows for more refined identification. (Check out the image below for the distinctions that LTU’s technology makes among different kinds of images.)
“We’re looking at the actual content of the image — we’re looking at the actual image,” Smith said.
Although more data-intensive on a frame-for-frame basis, LTU’s Image DNA system isn’t more costly that the current efforts to stamp or encode video for ultimate identification on the Internet, Smith argues. Content providers need only encode key frames of a video or TV show and then use web crawling tools to search the web, or scan specific suspicious sites, for content.
Still, hunting for copyrighted video on the Internet is bound to be a needle-and-haystack proposition given the vastness of the web and the staggering amount of copyrighted material that not only exists in archives but is also being created anew each day. Even given the uphill battle, LTU is getting the attention of Hollywood and the TV industry.
Smith says that his company has begun “a series of dialogs” with these content communities, primarily through their trade associations, the Motion Picture Association of America and the National Association of Broadcasters. Some of the discussions are well along and Smith believes that these groups and their members are more interested in using LTU’s technology to monetize content rather than yank content off the web.
(This item originally appeared in yesterday’s IP Media Monitor.)
Cynthia Brumfield at 9:07 AM|Comments(0)
Blogs We Read
