How is Google scanning thousands of books for its book search product? Of course, the company has developed a wonky, proprietary scanning method, and has patented it. That’s patent 7508978, “Detection of grooves in scanned images,” invented by Francois-Marie Lefevere and Marin Saric and assigned to Google.
The key is flattening pages to make them easier to process via word-recognition software. So instead of manually pushing books against a glass plate or slicing the binding off, Google uses an infrared projector and stereo camera to determine the curvature of a book’s page and the position of its spine, so it can generate a “flatter” version of the photo. In the end, that seems to be less work, and much gentler on the book. Neat.