A new medical camera designed to take pictures of internal organs is about to revolutionize the way you take snapshots. Soon you'll be able to take simultaneous video and stills of events that your naked eye could never see.
The best part? The camera was invented by two Oxford medical researchers out of off-the-shelf parts.
Faced with the prospect of paying $50,000 for a scientific grade camera that could record an underwhelming 80x80 images at 2000 fps, Oxford researcher Dr. Gil Bub hacked together an array out of existing camera components and projector parts. He not only created a much lower cost camera, but one that can simultaneously record images and video, and store the resulting data in a single image file. Said Dr. Bub:
"Normally, I'd have to use one of these expensive very low res, high speed camera and a second high res slow camera to capture the data, split the light optically, and go to the trouble of aligning each pixel by pixel."
The technique was termed Temporal Pixel Multiplexing (TPM), and the group took a camera sensor and placed in front of it a Texas Instruments DMD micromirror array, which has millions of tiny mirrors that can be flipped thousands of times a second. With this the researchers could precisely control which pixels were being hit by light at which point in time.
From this, each pixel was unveiled in a specific sequence, over the entire length of the exposure. In the example above, a 36 pixel sensor would be exposed in four batches. By using the mirror to precisely control each area of the sensor, two things happened: a large high resolution image was created by using data from the entire sensor; and by stringing together the sequential pixels, a high speed movie was created. The example below, the still image was recorded at 25fps (1/25 of a second exposure), while the video was 400fps
In addition to creating a device capable of recording both images and video at once, the team figured out how to combine the files into a single still photo. Tucked away in the blurred motion trails of that image is the high speed content, embedded in the image (click here to see a full sized version). If you enlarge the high resolution area sufficiently, you can actually see them. Since these areas are meant to be fuzzy anyway, the frames can fit there without being visually distracting. The longer the video recording, the longer the still exposure, so more blurry areas in which to embed the data. This doesn't even take much computer crunching, as Bub says:
"One neat thing is that there's very little processing involved. The way to think about the blurred regions is that if you were taking a TPM picture of a stationary object, there isn't any blur, but at the same time there isn't any motion to capture. If something moves, then the amount of motion is proportional to the size of the blur (more motion=more blur). But theres no real processing involved, just reordering the pixels. "
The advantage to this system is that it's less expensive than the dedicated scientific hardware; and faster, higher resolution and functions at lower light levels than the consumer high speed cameras, like those by Casio. While this working model was only 1360 x 1024 pixels, the concept could easily be scaled up. Dr. Mark Pitter of Nottingham University is currently set to work on a CMOS version of the chip, which wouldn't require the mirror system, and could fit in a normal camera — with a working prototype expected within a year. Within the scientific world Cairn Research, a UK based scientific instrument manufacturer is looking into the technology.