That's some way cool image processing. They perform a 2D FFT on both images, then filter out the low frequencies of Albert and the high frequencies of Marilyn. They combine the spectra and turn the result back into an image with an IFFT.
I can look at the image without glasses and see a perfect fuzzy Marilyn. Or, with glasses and see Albert.
You could get philosophical about how people see completely different things depending on what level of detail they are capable of taking in. We really only see a tiny part of the world around us.