Selective image analysis for roughness characterisation

Atomic force microscopy images supply precise information on real surface topography, but an automated method allowing accurate description of roughness through an asperity based approach is necessary for tribological applications. The present paper describes an original selective image analysis algorithm enabling confident indexation and analysis of surface asperities. First, the original image undergoes noise filtering followed by edge detection and thresholding. In the resulting binary image, clusters of asperities are separated by a selective treatment depending on objects' size and convexity. Finally, a regrowth of identified objects is applied to ensure their accurate three-dimensional shape estimation using a second order polynomial least square approximation. The identification performance of the method is hereby verified on an artificial image and illustrated with surface images of sputtered gold thin film alone and with an additional layer of ruthenium.