Comment by lettergram

This honestly seems like a standard approach to smoothing for computer vision (not to say, this isn't different in a technical sense).