If you’ve ever watched the “Jason Bourne” series of movies, he constantly has the CIA, INTERPOL, and other organizations following his every move. One of the ways they did this was through face-recognition software which took video feed, processed it, and compared faces which appeared in the video feed against an archived database. While, in the movie, the software never failed to recognize a face, this is rarely the case in real life. Even in ideal conditions, the best face recognition software only works well when you can control factors such as resolution or face angle. However, Pablo Hennings-Yeomans’ work over at Carnegie Mellon University may just change all that.
The biggest problem with the technology stems from the proliferation of low-resolution cameras, like those found in cell phones or lower-end security systems. As many of the facial idiosyncrasies are only distinguishable at higher resolutions. This results in choosing to use lower resolution source images, or trying to inject pixels into the video feed stills through use of super-resolution algorithms. However, Hennings-Yeomans’ method uses super-resolution algorithms and intermediate algorithms in a way that makes the distinct features of a face stick out. Hennings-Yeomans and B. Vijaya Kumar of Carnegie Mellon, and Simon Baker of Microsoft Research are presenting a paper at the IEEE International Conference on Biometrics going on later this month, canvassing the improved results when compared to traditional methods.
One quote, taken from MIT’s Write-Up on the new technology, describes Hennings-Yeomans’ work best:
“The approach shows promise, says Pawan Sinha, a professor of brain and cognitive sciences at MIT. The problem of low-resolution images and video “is undoubtedly important and has not been adequately tackled by any of the commercial face-recognition systems that I know of,” he says. “Overall, I like the work.”
Ultimately, says Hennings-Yeomans, super-resolution algorithms still need to be improved, but he doesn’t think it would take too much work to apply his group’s approach to, say, a Web tool that searches YouTube videos. “You’re going to see face-recognition systems for image retrieval,” he says. “You’ll Google not by using text queries, but by giving an image.”
Pictures courtesy of: MIT’s Technology Review;