As several companies look to develop a better way to cull through text based on the meaning of the keywords — or a “Google killer” — a maker of image-based search technology is moving to expand its potential customer base.
LTU Technologies, whose Image-Seeker image-based engine has mostly been used by law enforcement agencies to thwart online child pornography and counterfeiting, says the tool can also be used by major media companies to protect their copyrighted materials.
Viacom Inc.’s $1 billion copyright infringement suit filed this month against YouTube parent Google Inc. is but one example of the increasing efforts of media conglomerates to protect copyrighted content. Television shows and other copyrighted material are increasingly showing up on YouTube-like sites, significantly increasing the need for image-based search engines, LTU officials noted.
P. Kevin Smith, LTU’s vice president of sales for North America, said media companies are showing increasing interest in using image-based search tools to scan social networking sites for unauthorized use of their content. LTU already counts Meredith Corp., a Des Moines-based media and broadcasting company, among its clients.
“When you’re dealing with tens of millions of images … you need a way to automate that process of evaluating the known nonpermissable content and focusing on very specific filtering criteria,” Smith said.
A media company can provide LTU with video content, and LTU’s technology can monitor social networking sites to find out if they are using it, he added.
Image-Seeker is modeled after the human eye and how it and the brain work to understand information that is seen, Smith said.
The technology takes an image and builds “DNA,” or a binary file identifier of content, within the image. The identifier is based on attributes within the image, such as color, shape, texture and movement, Smith said. The tool uses this DNA to scan images on Web sites and flag those that may be using copyrighted material, exposed flesh or photos containing certain body parts.
“We are comparing that individual reference’s DNA to a profile that has been built … of 50,000 images commonly identified as porn,” Smith said. “The algorithms have been tuned to identify the content that is represented in each of those 50,000 images. We can feed back images that are consistently being missed as false positive or false negatives to train the algorithm … so the images will be caught in the future.”
Last November LTU released a new feature for its search tool to allow law enforcement officials to scan in real time a hard drive obtained in the field for suspected pornography, Smith added. “It really shortens the time to get into more difficult spaces on a hard drive like unallocated space,” he said. “Instead of a series of days to get into that area, law enforcement has been able to do it in hours.”
The FBI, the Bureau of Immigration and Customs Enforcement, Interpol, the Royal Canadian Mounted Police and other organizations are among LTU’s clients.