Signal-to-noise ratio estimation technique for SEM image using B-spline

Abstract

A new signal-to-noise ratio (SNR) estimation technique is proposed for the scanning electron microscope image. Based on only a single image, an estimation technique named B-spline is proposed. Three existing techniques are applied to compare with the performance of the proposed method in terms of the zero-offset point, SNR and percentage error. They are the nearest neighborhood, first order interpolation, and a combination of these two methods. A t-test is conducted on the proposed method. Noise variance is estimated from the SNR calculated and a Wiener filter will be used to filter the noise with the filtered images being similar to noise-free images.