Disney Research (an unofficial joint effort with labs such as ETH Zurich to analyze and develop for companies like ESPN, Pixar Animation Studios, Walt Disney Animation Studios, and so forth) has developed a breakthrough which can induce 3D computer models from 2D images.
Not only that, the quality of these 3D computer models is substantial to meet the needs of makers of films and video games as well, Disney Research has claimed. The algorithm used to generate these models require multiple images to capture the scene from a range of vantage points. However, the 3D model has its restrictions, as it is only intelligible within the field of view surrounding the original images.
Having said that, the detail achieved (as you can see from Disney Research's demo video) is still pretty remarkable. Unique from existing systems, this algorithm measures the depth for every pixel, so it's especially effective at the edges of objects. But the system is not flawless yet. Depth measurements are not as accurate as they would be when captured with a laser scanner, and more effort has to go into handling surfaces that vary in reflectance.