When I saw Photosynth for the first time about two years ago, it joined a small handful of new products that really captured my attention. The software arranges sets of photos in 3-D context and allows viewers to navigate fluidly from image to image, moving their gaze from a building’s facade to a detail shot of a specific fresco, for example.
Photosynth is a distinctly Seattle invention. It emerged from a collaboration of University of Washington graduate student Noah Snavely and computer-science professor Steven Seitz, with Microsoft researcher Richard Szeliski, as well as a Ballard startup Microsoft acquired. Now at least part of that team is at it again.
In a paper presented at this week’s SIGGRAPH (a meeting of the world’s top computer graphics researchers), the UW/Microsoft team described the next iteration of their work, soberly named “Finding Paths through the World’s Photos.” Here’s the video:
As the video explains, they’ve improved the photo navigating experience, making it smoother and more natural. If you want to zoom from a view of a building’s exterior to its interior, the system navigates a path through other photos of the scene to let you glide in through the building’s door and gradually turn your gaze through several images until you reach the one you want. As the video demonstrates, it’s a lot like what you’d see if you walked inside and turned your head.
Other improvements include color balancing, stabilization and the ability to smoothly pan and orbit around an object.
There’s also a way to embed your own photos in broader collections. So if you have some snaps of your family in front of Notre Dame, for example, the system could put the images in their 3-D context and use the smooth navigation to travel from one to the other, filling in the space between your photos with others in the broader collection.
It’s easy to imagine the possible applications for this technology. Microsoft’s chief futurist Craig Mundie described scenarios at the company’s Financial Analyst Meeting last month that would seem to be a natural fit:
“So how is the Web going to evolve? Well, it’s gone from text to pictures to animations to video to a sort of 3-D immersive environment. We’ve brought you Virtual Earth and things like that. … And the question is, where do we go next? And I think the next thing we’re gonna do, which will complement this natural way of interacting, is what I’ll call the visual Web. …
“We think that the idea of first life [as opposed to the made-up world of Second Life], where there’s a mirror world of 3-D that everybody can participate in constructing and maintaining and which gives us a navigational metaphor that’s completely consistent with the world we already live in would allow many more people to get into this environment and operate there.
“So, in fact, I think we’ve shown before some Photosynth technologies that the company perfected where we have many pictures that can be contributed on the Internet by professionals or cars driving by or people. And what we have is a server-based technology that composites these and creates a 3-D world.”