Microsoft pictures a 10-gigapixel photo
He's already done 4-gigapixel shots of downtown Seattle.
Cohen's work, dubbed Big Panoramas, is an attempt to marry Internet mapping and high-resolution photography. With 4 billion or 10 billion pixels, a single photograph will contain several square miles of real estate in accurate detail. In the Seattle photo, users can zoom in on windows on different buildings, or zoom out to get a view of the entire skyline.
The end result is something akin to the satellite images on services like Google Earth. The difference is that the angle is more familiar. The pictures provide the panorama you might see staring out of a window on a building, or from standing on the sidewalk. Satellite images capture only the unfamiliar bird's-eye views of rooftops.
Ultimately, several-gigapixel shots captured from different angles could be woven together to form a 3D-like photograph consisting of tens of billions of pixels, Cohen said.
The technique involves taking several hundred pictures with a standard digital camera, stitching the photos together and then compensating for changes in the position of the sun, the movement of clouds and other environmental factors during the time it took to take all of the photos.
In 2004, engineers in the Netherlands stitched together 600 individual images shot over the course of more than an hour to create a photo with almost 2.5 billion pixels.
The 4-gigapixel photo of downtown Seattle required shooting more than 800 photos taken in an hour and a half. The lighting and different exposure conditions that existed during that time period, however, are neutralized so that it looks like the entire image was captured at a single moment.
"With a 10-megapixel camera, a 10-gigapixel picture takes at least 1,000 pictures," he noted.
The camera is not held by a person. It sits in a motorized rig and the angle of the rig and camera are controlled by a computer.
Digital cameras aimed at consumers, by contrast, typically offer resolutions of about 5 or 6 megapixels.