科学美国人60秒:Fancy Figuring Ferrets Out Fake Photos

One of the most effective ways to spot less obvious fakes is to look at the shadows cast by objects in the picture. If they’re pointing in different directions, chances are the image you’re looking at isn’t real. But the analysis can be challenging when the objects have complex shapes and the shadows are cast on uneven surfaces.

That’s why Dartmouth and UC Berkeley researchers have developed software that analyzes images on two fronts. The program tracks the shapes and angles of different shadows for inconsistencies. It also uses geometric algorithms to determine the angle of illumination on each object, which should match up if there’s a single light source. This research appears in the journal ACM Transactions on Graphics. [Eric Kee, James O’Brien and Hany Farid, Exposing Photo Manipulation with Inconsistent Shadows]

The researchers tested their software on a 1969 moon-landing photo. Conspiracy theorists often point to the complex orientation of shadows in the picture as evidence it was doctored. But the computer analysis says it’s kosher. Looks like we might have made it to the moon after all.

—Larry Greenemeier

[The above text is a transcript of this podcast]