statistical and spectral texture features of the SEM image

A study of methods for texture classification of SEM images of micro-surfaces of objects and their segmentation

Purpose. The goal of this work was to develop and study the methods of texture classification of SEM images of micro surfaces of objects based on the statistical and spectral characteristics of texture fragments, as well as a comparative analysis of segmentation methods of SEM images. Methods. The determination of the texture characteristics was based on statistical moments computed by the brightness histogram of a SEM- image or its region. The spectral measures of texture of SEM image were based on properties of the Fourier spectrum.