This is a fantastic video. If I understood the explanation correctly, the algorithm applies a multi-frame low-pass filter for color and position, and frame-by-frame exaggerates the delta from that baseline. The resulting video greatly amplifies small changes in color or position revealing small but significant variability in the scene.
Originally shared by Jason Gullickson
Pete Prodoehl I expect a Processing implementation of this by the end of the week 😉
http://bits.blogs.nytimes.com/2013/02/27/scientists-uncover-invisible-motion-in-video/
http://bits.blogs.nytimes.com/2013/02/27/scientists-uncover-invisible-motion-in-video/