It is filtering, the input (video) is filtered down to specific features. i.e. in the cat filter it is filtering everything except eye position, mouth and maybe chin position relative to eyes.