An Algorithm Found These Images Hidden in Reflections From Glitter

By Sarah Zhang on at

There's a new image algorithm called SparkleVision in town, and yes, it involves glitter. A glitter-coated surface, you see, is really just a plane of many, many tiny mirrors. And a shattered glitter-reflection can indeed be reconstructed into the face of Obama.

Before there was SparkleVision, there was a glitter board. A trio of MIT scientists started by reflecting images on a computer screen off the board and photographing the reflection. To the human eye, the glitter reflection is just a shimmery surface. But each piece of glitter is also a small, randomly oriented mirror, what the scientists wonderfully call a "random specular microfacet" in their paper uploaded to ArXiv.

By comparing several images before and after reflection, they were able to create an algorithm that mapped each pixel of the original image to its reflection in glitter. So here you have it, Marilyn Monroe and Albert Einstein emerging out of shimmery nonsense.

An Algorithm Found These Images Hidden in Reflections From Glitter

But there's more to SparkleVision then shits and giggles. While this initial iteration of SparkleVision is limited to the exact set-up used in the paper, the researchers hope to eventually apply it to forensic reconstruction of reflections from other similarly shiny objects—such as certain metals, fabrics, and wet surfaces. [MIT Technology Review, ArXiv]

Images from Zhang et al.