Graham Flint takes really big pictures. Really, really big pictures. My digital camera captures 4 megapixels – Graham produces digital images that are 1000 times that size – 4 gigapixels. What this means is that Graham can start his talk at Pop!Tech with a distant landscape shot of
San Diego, shot from 3 kilometers away. Graham selects one percent of that image and zooms in – there’s no meaningful blurring or distortion. So he zooms in again. Now we’re looking inside a hotel room – from 3 kilometers away – and we can make out the bed, the paintings on the walls, the patterns on the curtains.
Yep, that’s the image. It’s high enough quality that you can look into each of those hotel rooms…
Oddly enough, the way to make really huge digital images is to shoot film. Really big film. Why would you make digital images using film? Since each image is shot in as little as 1/200th of a second, the data transfer rate you would need to capture these images would be 38 terabits per second – that’s an amazingly high data rate you’d need to capture – without massively parallel computation, there’s no way to handle that much data in that short a time.
So Flint has build a film camera. (Indeed, it’s a really, really big camera.) It uses film magazines salvaged from U2 spy planes (Flint used to run one of Lockheed Martin’s laser labs, which gave him access to some interesting technologyy.) It shoots 460mm x 230mm film stock using lenses that are anywhere from 200mm to 500mm in length. Those lengths would usually be telephoto lenses – but with film this big, these lenses act like wide angles, letting Flint photograph landscapes from 10-20km away. No commercial lenses are sufficient for this work – he and his team grind their own lenses, made of six different types of glass, and custom fit them to 30 kilogram cameras. The sheer geekery required to build these cameras is astounding – and the geekery to take a shot (laser rangefinders, adjustment screws that are tuneable to a thousandth of a centimeter…) is profound as well. And then scanning and digitizing the picture involves hours, terabytes of storage, and lots and lots of touchups in Photoshop.
So why the heck is he doing this? One is that he wants to create real, compelling virtual reality. To give a hemisphere of visual information (as you might get in an IMAX theater) at 20/20 vision, you need 75 megapixels. Add a 10x zoom in any direction and you need 7.5 gigapixels… and the cameras Flint is building aren’t quite adequate yet.
Flint and his wife have been travelling around the US and Canada, shooting a picture or two a day, for a “warmup project”, Portrait of America. In three years, they’ve got 1,300 (very very big) pictures. The long term project is a portrait of the world, being put together in cooperation with the CTO of Google Earth.
There’s some disturbing implications of this technology, though – Flint shows us a beautiful photo of a paraglider over a stretch of beach. As we zoom in, we discover it’s a nude beach, and Flint has photographed a bunch of nude bathers from over a mile away. Not his intention, but certainly a consequence of being able to take really big photos.