The University of Washington, who, along with Microsoft Live Labs, gave us the 3D photo-browsing technology that is now called Photosynth, has come up with a new algorithm that could expand on Photosynth’s capabilities. In fact, it takes the 3D rending of Photosynth and uses it to model entire cities using tens or hundreds of thousands of photos. To demonstrate what the new algorithm can do, the researchers used the search term “Rome” on the digital photo sharing site Flickr. The results returned more than 2 million photos. In this group, every popular landmark, facade, interior, fountain, sculpture, painting, cafe, etc. was documented.
With the new algorithm, the researchers were able to reconstruct the city using the photos downloaded from the site. To do so, it matches common points found in the images. The process unbelievably only took 21 hours.
To see the technology in action take a look at this model (see video below) of the Colosseum constructed using 2106 photos. In the video, you’re seeing the overall view, but you would normally be able to zoom into individual images too.
Besides Rome, the researchers are also working on constructing models of Dubrovnik, Croatia and Venice, Italy. They plan on presenting their paper at the International Conference on Computer Vision in Kyoto, Japan next month.
For more video demos, check out this page.
(via TechFlash)
