Ramesh Raskar's team at the MIT Media Lab is re-imagining what photography can do
By creating movies of light in motion at a trillion frames per second, camera can see around corner
Raskar: Technology is years away from applications in the real world
He says it could be helpful in preventing car crashes and aiding rescues
Editor’s Note: Ramesh Raskar is Associate Professor at MIT Media Lab and head of the Camera Culture research group. He is an Alfred P. Sloan Research Fellow (2009), a recipient of the DARPA Young Faculty award, the co-author of “Spatial Augmented Reality“, and the holder of more than 50 US patents. He spoke at the TEDGlobal conference in June. TED is a nonprofit dedicated to “Ideas worth spreading” which it makes available through talks posted on its website.
A camera that can see around a corner?
I know this sounds like something in a sci-fi movie or a superhero comic, but this is a real-world technology we’ve made possible with a camera that is aware of the travel time of light, an imaging technique that can create movies of light in motion with an effective rate approaching a trillion frames per second: the speed of light.
Before I joined the MIT faculty in 2008, I had done deep research in “computational photography,” a field of new imaging techniques dramatically improving the capture and synthesis of photos. But, I knew there was more to photography than just depicting what the eye can see. I wanted to create a camera that could see beyond the line of sight. The speed of light isn’t infinite: light travels about a foot per billionth of a second.
If I could build a camera fast enough to analyze light at high speeds in room-sized environments, I knew we could then create cameras to solve major problems in scientific and consumer imaging, and enable completely new functionality.
I spoke to top researchers in ultrafast lasers and photonics to understand what was currently possible. When I did, most of them asked some version of: “Why? Why spend years building a camera to look around corners when no commercial application is screaming for it and no funding agency has a call for it?” In addition, it’s rare to shoot light pulses and analyze at such high speeds in large environments. Ultra-fast imaging experiments are usually limited to centimeter- or smaller-size samples.
I continued the work and in the spring of 2008, with James Davis from UC Santa Cruz, wrote a proposal that laid out the mathematical foundation and various experimental solutions for exploration.
I also began working with Media Lab graduate student Matthew Hirsch to build a prototype, hoping that we would have something to demonstrate within a few months. The grant proposal was rejected for administrative reasons (I made a paperwork error!), which meant we had to wait nearly a year to apply again. But those two years didn’t yield any meaningful results, as our lab components weren’t designed to be used the way we wanted.
After nearly three years of experimental work, the team – especially postdoctoral associate Andreas Velten and MIT professor Moungi Bawendi, many students, and several collaborators – cobbled together pieces of the puzzle and built a software program to create a first demonstration of looking around corners. Very soon afterward, we could also start creating surreal movies of light in motion.
We call this new imaging technology femto photography because we capture a segment of the image with a flashlight (in this case, a laser pulse) on for a few millionths of a billionth of a second (or a few femtoseconds) and an exposure time approaching a trillionth of a second.
Just how fast is femto photography? Think of it this way: if we took one-thousandth of a second of footage from the femto camera video output and slowed it down to the speed of 30 seconds per frame – the approximate speed of a standard TV broadcast – it would take us a lifetime to watch.
Photographers know that at very short exposures and even at the most sensitive setting for dark scenes, we will record barely any light. So what about in a trillionth of a second? We actually record and average millions of photos to get enough light, each photo made to look the same via carefully timed synchronization with the light pulse. So even if our exposure time is indeed nearly a trillionth of a second, to get sufficient light we must take an average. Thus, as of now, we can only record repeatable events, but this is not a fundamental limitation.
Unlike conventional cameras, our femto camera captures an image as one thin slice at a time of one-dimensional space using a “streak tube,” a laboratory instrument that is commonly used by chemists to study light passing through chemical samples. We then take hundreds of these narrow videos (each shot at a slightly different angle) and create a carefully synchronized, slow-motion composite of the light pulses. It takes about an hour to collect and aggregate the data (view a demonstration of a light pulse as it travels through an ordinary Coca-Cola bottle).
To see around the corner, we use femto photography to analyze scattered light. We bounce light off of visible parts into hidden parts and then measure the time and direction of returned light.
Usually the scattering of energy is considered a nuisance – whether driving in fog or poor reception from a cell phone tower – and most techniques either try to avoid it (by turning on fog lights) or reduce the impact of scattering (by selecting energy for the phone only from direct paths).
In contrast, we exploit the scattering. For the camera, a laser pulse is fired at a wall, and the impact of hitting the wall causes the particles of light to scatter. Some of the scattering particles return to the camera at different times. This is repeated about 60 times per image as the camera measures how long it takes for the light to travel back and where the particles land. An algorithm then crunches the data to reconstruct the hidden image. This technique even allows us to see a three-dimensional object such as a mock-up of running person.
As exciting as this work is, don’t look for this technology to be in practice tomorrow – we’re still years away from bringing this to market. But, we can already imagine multiple ways that it could have a significant, positive impact on our everyday lives.
By allowing us to “see” around a corner, for example, this technology added to our cars could let us know if there’s another vehicle approaching around a blind curve. It also could give us a new way to look deep inside our bodies without X-rays, or to look through a window into a burning building from a distance to see if anyone is left inside –without risking a firefighter’s life.
When I gave a TEDGlobal talk on femto photography in June, I began with a reference to Doc Edgerton, a very popular MIT professor of electrical engineering who, in 1964, wowed the world with an image of a bullet in midair, having just passed through an apple. He accomplished this by using a strobe light to freeze the action of the bullet at a millionth of a second.
What we’re talking about here – the speed of light –is a million times faster, and is opening the door to a complete rethinking of what we mean by, and can do with, photography. It is a first step toward a new world of imaging that far exceeds human ability, resynthesizing data and depicting it in ways that are within human comprehensibility.
Follow @CNNOpinion on Twitter
Join us at Facebook/CNNOpinion
The opinions expressed in this commentary are solely those of the author, Ramesh Raskar, who would like to thank his collaborators: Moungi G. Bawendi, Professor, MIT Department of Chemistry; Andreas Velten and Christopher Barsi, postdoctoral associates, MIT Media Lab; Everett Lawson, Otkrist Gupta, Nikhil Naik, Amy Fritz, Di Wu, MIT Media Lab; Matt O’toole, Kyros Kutulakos, University of Toronto; Diego Gutierrez, Belen Masia, and Elisa Amoros, Universidad de Zaragoza; Kavita Bala, Shuang Zhao, Cornell University; Ashok Veeraraghavan, Rice University.; Thomas Willwacher, Harvard University.