Efficiency Evaluation of Segmentation Algorithms for AFM Images

Abstract

The results of evaluating the efficiency of algorithms for segmentation of images of surfaces of materials with an absent or weakly expressed substrate and a convex shape of objects obtained using an atomic force microscope (AFM images), as well as synthesized in the Matlab and Gwyddion software are presented. For segmentation, algorithms were used based on wave growth of local maximum regions with their selection in decreasing order of values (without stopping and with stopping at a given level), marker watershed (with automatic placement of markers, under the control of the operator), watershed based on distances, growing areas (without selecting starting points, with the choice of starting points based on extrema), the Vincent – Sulli watershed (classical, with a preliminary calculation of the gradient in an eight-connected area, with the selection of the contours of the areas and their subsequent filling), a two-phase watershed. Segmentation algorithms realization in Matlab and in the specialized software package Gwyddion are considered. Algorithms efficiency was assessed using segments number, uniformity brightness within a segment, contrast at the border of adjacent segments, and a complex criterion that takes into account the uniformity of segments brightness, their number and size.