Laura S. Marshall, firstname.lastname@example.org
There’s one big trend in forensic video analysis that has experts crying foul, and that’s the increasing shift toward digital media.
“Video has kind of overrun where it should have been,” said Alan Salmon, a forensic video analyst for the Oklahoma State Bureau of Investigation. Salmon is the executive vice president of the national Law Enforcement and Emergency Services Video Association (LEVA) and will become the group’s president this month. “VHS was kind of done away with way too quickly.
“If you have two hours of tape – 140 gigs of information – and you put that two hours on a 4.7-gig DVD, think about how much information you’re throwing away.”
In a field such as forensics, information can mean the difference between conviction and acquittal: A dimple or a scar could be used to identify a suspect, so having it appear on the analog tape and disappear in conversion to DVD could spell disaster for prosecutors.
Analyzing digital video
Whether dealing with digital or analog video, there are certain functions that video analysts need from software systems: They might have to brighten a dark scene or clean up a picture that is too grainy to be useful, and there are a number of programs that can help with those tasks and more.
For digital video – even from sources such as cell phones or YouTube – there is Ikena Reveal from MotionDSP Inc. of San Mateo, Calif.
Other programs rely on frame averaging, said Sean Varah, CEO of MotionDSP. If you take 30 frames of video and line them up just right, he explained, you can pull a license plate number from what originally was just a fuzzy parking garage security tape.
But it takes time: First there is the lining-up stage, then the stabilizing stage, then the sharpening stage and maybe brightening or other enhancement.
“You might do it in five separate steps,” Varah said. “Reveal automates all that.” It speeds up the process, too, he added.
A key strength lies in what Varah calls complex-motion scenes. “Frame averaging only works with a single type of motion, like a car moving past a fixed camera. It falls apart when you have complex motion, where both the camera and the object are moving, like filming a car driving by with a handheld mobile phone – both the car and the camera are moving. This is where we excel.”
Keeping things interesting on the analog side is Cognitech Inc., which offers a suite of analog-video analysis programs made up of Video Investigator, Video Active and Auto Measure.
“The CIA uses it – all over the world, people use it,” said Leonid Rudin, CEO of the Pasadena, Calif.-based Cognitech. He also serves as professor of imaging science at École Normale Supérieure in Cachan, France, and as chairman of the investigative/forensic imaging conference group at SPIE.
“Photoshop on steriods”
Video Investigator, Rudin said, has been around since the 1990s. He likened it to “Photoshop on steroids, with over a hundred plug-ins based on algorithms.” Video Active was designed to allow investigators to do things on the fly. “If you’re investigating videotapes, you don’t have time to go in a lab and look through every single frame,” he said. “So we came up with software that runs as fast as videotape.” Auto Measure applies Rudin’s special interest in 3-D to videogrammetry, or to measuring objects and distances in video.
Not content to rest on their laurels, Rudin and the other researchers at Cognitech are looking to add 3-D imaging and identification features to their products.
“There are two directions things will likely take,” said George Reis, president of Imaging Forensics in Fountain Valley, Calif. “One is that small companies will create proprietary-based solutions that seem great but have too small a market to support, and will be too costly for most forensic users to take advantage of. The other is that standards will eventually be implemented – some good, some will miss the mark, but we’ll hopefully be able to share data between systems and get better tools.”