Fast filtering of surface topography of micro-optics using non-local means

Abstract

Micro optical components are more and more widely used in precision engineering due to their small sizes and novel functionalities. Characterization of the surface topography of these components is very difficult due to the existence of sharp edges and complex features. Conventional filtering algorithms cannot be used directly for non-smooth structured surfaces. In this paper we present a filtering algorithm using the non-local means method. Instead of assigning weights according to the closeness or similarity between individual data points, this method are based on the similarity of the patches surrounding data points. This method can effectively separate the detailed textures of non-smooth surfaces while preserving primary features. Proper adaptation and improvement are made for the applications in precision engineering. The k-means clustering method is used to reduce the computational cost. Numerical experiments prove that the non-local means method is able to separate small-scaled textures from the primary surface shapes without ruining the sharp features.