NASA scientists David Hathaway and Paul Meyer weren’t looking at the sun and studying satellite maps to help fight crime. But somehow the work they were doing led the FBI to their lab at the Marshall Space Flight Center in Huntsville, Ala.
It was 1996, and the federal agents needed some help with a high-profile case, in which some homemade video could prove very helpful. Hathaway and Meyer, who had worked a lot with image processing techniques, stepped forward to help, and their foray into crime-fighting began with the bombing of Centennial Park during the ’96 Atlanta Olympics.
At first it might seem surprising that the FBI would turn to NASA for this type of help, but because the job involved deciphering images, who better to turn to than the creators of the Hubble Space Telescope?
Following a Pattern
In their day jobs with the national space agency, both men have to routinely distill information caught, but often hidden, on film, and for years they had been refining ways to overcome the limits of poor video footage.
“With telescopes on the ground used for looking at the sun, there’s always some jittering,” says solar astronomer Hathaway, who had developed stabilization methods to minimize shaky images.
Meyer, experiencing similar problems as a meteorologist, had to make sense out of satellite pictures that were too jumpy, too blurry or too noisy. If you’ve ever seen a freeze frame, you’ve seen video noise. It’s the speckly colored pattern that always looks worse than the moving video itself.
So the men began working on what would become one of their greatest challenges at NASA — getting a clearer picture from moving, shaking, amateur video footage that could help solve a crime.
The Making of VISAR — It All Adds Up
Both men say it took a lot of trial and error, but eventually they came up with a successful way to stabilize a rough image. They call their technique the Video Image Stabilization and Registration, or VISAR for short. And while it seems nothing short of miraculous on the surface, underneath it all adds up — quite literally.
To get something from nothing is impossible, but to get something from an accumulation of clues works quite well.
Meyer uses a simple analogy to help picture how the system translates a good image from a poor one.
Think of three transparencies each containing a colored square, triangle and circle. If you take two of these and drop them on top of each other, the shapes don’t necessarily match up. But getting them to match is the goal. So you shuffle and move the slide around, left, right, up, down, to get the best possible match between these shapes. Then you start rotating them and possibly even zooming in and out, so that the the square’s on the square, the triangle’s on the triangle and the circle’s on the circle.
“Each of these transparencies is a video frame, and there’s thirty of those per second,” says Meyer. By adding these thirty pictures every second, the more footage you have the clearer and more stable picture becomes.
Getting Rid of the Noise
Hathaway explains that it all starts with a single frame. The VISAR user identifies exactly what parts of the video frame he wants to make clearer — choosing the real-world square, triangle and circle from Meyer’s explanation.