Researchers have devised a way to read books when they’re closed. No, you didn’t read it wrong. Researchers from MIT and Georgia Tech are designing an imaging system that can read closed books.

A paper published Friday in the journal Nature Communications describes a prototype for this ingenious system that correctly identified the letters on the top nine sheets of a stack in which each sheet had one letter printed on it.

“The Metropolitan Museum in New York showed a lot of interest in this, because they want to, for example, look into some antique books that they don’t even want to touch,” Barmak Heshmat, co-author of the paper and a research scientist at the MIT Media Lab, said in a statement. He added that the new system can analyze materials organized in thin layers.

The system uses terahertz radiation, which unlike X-Rays, can distinguish between ink and paper. It also gives better depth resolution when compared to ultrasound. Paper and ink bend light at a different degree, which helps the system in distinguishing between them. The 20-micron deep air pockets between the pages help the system differentiate between the pages of a book.

Most of the radiation is either absorbed or reflected by the book, some bounces between pages and returns to the sensor producing a false signal. The sensor also produces a background hum, all of which the researchers have to filter out.

MIT researchers developed algorithms that receive images from individual sheets and researchers from Georgia Tech developed algorithms that interpret distorted or incomplete images.

After determining the distances between the first 20 pages of the book, the system could pick out the individual letters printed on the first nine pages. With the help of femtophotography, which can capture certain types of images trillionths of seconds apart, the system could make out if the image came from a certain page or the next one below it.

Researchers are now working on improving the accuracy of the detectors and the power of the radiation sources so the system can identify images beyond nine pages.