Comment by tchadwick

I assume they are using the theory/algorithm from the MIT video magnification work: http://people.csail.mit.edu/mrub/vidmag/