Virtual Photography

Back in 2001, CBS introduced EyeVision for their Superbowl telecast. EyeVision allowed the broadcasters to combine images from several dozen cameras, positioned and 7° intervals around the stadium, into a seamless playback that could be rotated on the fly, creating an effect much like the famous “Bullet Time” sequence in The Matrix. The technology, while not perfect, was really neat, and offered unparalleled insight into the on-field action. Unfortunately, it hasn’t made a reappearance at a Superbowl since then, and the technology doesn’t appear to have been pushed forward much.

Fast forward to 2011. The Kinect has been released for the XBox 360 and has been enthusiastically embraced by hackers. Microsoft, while initially reluctant to let people fool around with the sensor, quickly realized that they were fighting a losing battle, reversed their position, and now even provides some official support for people doing new and interesting things with it.

The Kinect works much like a WebCam: you point it at something, and it gives you a video feed of what you’re pointing at. The interesting thing, however, is that in addition to the image, it also tells you how far away each element in the image is. Using a couple of Kinects in tandem allows one to actually reconstruct a texture-mapped scene in three dimensions. Once you’ve done this, it’s a simple matter to move around and through the space, much like you would in a video game. Here’s the first proof-of-concept I saw:

And here’s another demonstration that uses a single Kinect panning around a scene to capture the data:

As you can see, once the data is assembled, one can move a virtual camera through the scene, viewing it from angles where there was no camera to begin with. Though these first efforts are still a bit rough, think what one could do with this sort of technology if it were refined and made more sensitive: the Superbowl’s EyeVision technology could be expanded: instead of 33 possible vantage points, you could see a playback from an infinite number of angles, even swooping in among the players. Movies could be filmed with full 3D data sets, allowing one to move through a scene and see it from whatever angles one wished. Professional photographers could not only adjust exposure, contrast, saturation and that sort of thing, but also the apparent angle from which a photo was taken.

I expect this sort of thing to blossom over the next few years, and am anxious to see what happens with it once it gets from the hands of hobbyist technologists to those of artists and producers. It should be a fun ride.