Review of single image signal-to-noise ratio estimation for SEM image

Abstract

Scanning electron microscope (SEM) image signal-to-noise ratio (SNR) depends on the beam current, the materials present in the specimen, and the specimen topography. It is desirable to quantify the SNR in SEM images, as it is a parameter, along with spatial resolution, that quantifies the image quality. SNR measurement usually requires at least two images, to avoid this requirement, a method of SNR estimation with only a single image is described here. The SNR could be quantified as the ratio of signal variance to noise variance. The autocorrelation of image at its peak (zero offset) is used to estimate the noise variance and the signal component in accordance to the corresponding original autocorrelation and mean of the image while assuming the signal and the noise are uncorrelated.