‘Computational Zoom’ Lets You Remix Focal Lengths for Impossible Photos

Traditional photography involves choosing a focal length and position before snapping a shot. A new technology called “computational zoom” will allow photographers to remix those two factors afterward to create not only new compositions, but physically impossible ones.

The UCSB Current reports that the project is being worked on by researchers at NVIDIA and UCSB. Here’s a 5-minute video that introduces and explains “computational zoom”:

Here’s a short explanation of how it works. When composing a photo, your focal length and position can drastically alter what the resulting image features. In an example given by researchers, photographing a woman with a 16mm lens up close helps to reveal a roller coaster looming in the background, while photographing her from a distance using a 105mm lens focuses on the train bridge far behind her.

What’s more, her body shows more distortion up close with a 16mm than when captured afar with a 105mm (keeping her body the same size in the frame) — it’s how a camera adds 10 pounds.

But what if you want to have her body from the 105mm shot but the roller coaster from the 16mm shot? That’s now easy with “computational zoom” technology. The framework allows you to split up a scene based on depth, and assign a different focal length perspective to each of those depths. You can make the foreground look like it was shot with a telephoto lens and the background look like you used a wide-angle one.

First, you need to shoot a “stack” of photos with a fixed focal length. Starting from a distance, you move closer to your subject with each new shot.

The stack is then inputted to the “computational zoom” framework, which calculates the structure of the scene using the photos, creates a 3D reconstruction, and then allows you to render the scene from multiple perspectives. In this example, you can zoom “out” on the roller coaster while keeping the woman unchanged:

Here are some short animated GIFs showing what’s possible with remixing focal lengths in this way (these multi-focal length photos can’t actually be captured with a single exposure and a single lens):

“This new framework really empowers photographers by giving them much more flexibility later on to compose their desired shot,” says researcher Pradeep Sen. “It allows them to tell the story they want to tell.”

“Computational Zoom is a powerful technique to create compelling images,” says NVIDIA scientist Orazio Gallo. “Photographers can manipulate a composition in real time, developing plausible images that cannot be captured with a physical camera.”

The researchers are hoping that computational zoom will one day be offered to the general public as a plug-in for popular image-processing programs.