Richard Szeliski, principal researcher at Microsoft Research who developed the technology with Noah Snavely and Steven Seitz, of the University of Washington, said: "The system builds a 3D model just from a raw collection of photographs."

'Geometry problem'

He said the technology works by scanning collections of photographs, which can be taken from different cameras at different times.

It picks out distinctive features in each image and cross-references them against the other photographs, checking for similarities.

This allows it to pinpoint a feature's 3D position and to also calculate where the position of the camera would have been when the picture was taken.

"Then basically, it is just a geometry problem," explained Dr Szeliski.

"You are simultaneously adjusting the position of the camera and where those little pieces of images are until everything basically snaps together."

The prototype can use as few as two pictures but, said Dr Szeliski, it becomes a lot more exciting when several dozen images are used.

The 3D model will enable people to move through a scene, looking at pictures from any angle, click on specific photos, zoom in or out of a feature, and see where snaps were taken in relation to other images.

Dr Szeliski said: "I think the photo-sharing websites will be early-adopters of this technology.

"Wherever people share photos, instead of just seeing a gallery of unorganised photos, now you can pull everyone else's photos together and make a rational sense out of it."

The other obvious application, he added, would be for tourism and property, where a city could provide a virtual tour or a hotel could potential visitors walk through its lobby.

The researchers will be presenting a research paper detailing the technology at this week's Siggraph 2006 conference in Boston, which looks at the latest developments in computer graphics and interactive techniques.

Microsoft have said they believe the technology will almost certainly be web-based, and people should be able to run the application later this year.

"Within the next year we hope to see this in wide usage," added Dr Szeliski.