Reconstruction of multi-frame semi-sparse scanning probe microscopy images using dependent Gaussian process

Abstract

A major drawback of scanning probe microscopes (SPMs) is the time-consuming image acquisition, typically on the order of tens of seconds to minutes. Sparse raster scanning and non-raster scanning have been demonstrated to be effective approaches to improve the efficiency of SPMs. However, these strategies entail the challenge of reconstructing accurate images from semi-sparse sampling points, which means the distribution of the scattered sampling points is considerably sparser in the direction perpendicular to the scanning path. In this paper, we focus on the multi-frame semi-sparse video image reconstruction. In this setting, one of the frames is regarded as the reference frame while the others are regarded as the target frames. Instead of reconstructing the target frames independently, we propose a dependent Gaussian process regression model to make full use of the information provided by the reference frame and the currently processed target frame. The proposed method takes advantage of the intrinsic correlation between images. Thus, the accuracy of the reconstruction can be improved greatly. Comparisons with the renowned Delaunay triangulation-based interpolation method and the Gaussian process regression method prove that the proposed method has dramatically improved the reconstruction accuracy of semi-sparse video images.

Feature correlation method for image reconstruction evaluation in under-sampled scanning probe microscopy

“…This technique involves a scanning pattern that either bypasses high-frequency signals [11][12][13][14][15][16][17][18] or reduces scanning points, followed by image reconstruction through software. [19][20][21][22][23][24] Its advantages include fewer surface scans, reduced scanning time, and a shorter total measurement duration. Under-sampled SPM allows a short time for SPM measurements, but it is necessary to evaluate whether the data acquisition and reconstruction are performed correctly because the data at unacquired positions are compensated for by estimated values.…”