Going a tiny bit farther than “optical illusion” (which is right on the mark):

Color vision works best only near the absolute center of the eye, the further from that the more color blocking the brain does.

Your brain is attempting to turn the static images into a more continuous stream, thus features from the previous picture/s can distort those that come after.

Your brain actually stores memory of faces as separate components and there seems to be a critical junction in the brain that integrates these features so that so can recognize faces as unique. By not looking directly at the images you are getting an interpretation of those parts more so than recognition of the full figure.

There are a lot of fun hiccups in the brain that makes this work but those are a few items that could be the basis of this individually or in combination.