Reconstruction of undersampled atomic force microscope images using block-based compressive sensing

Abstract

Atomic force microscope (AFM) is a high-precision instrument to research surface topography of various samples. Nevertheless, the standard AFM procedure takes excessively long time to acquire high-resolution images. Moreover, too much interaction between a probe tip and a specimen potentially leads to tip abrasion and sample damage. Compressive sensing (CS) has been employed to realize high-speed AFM (HS-AFM). However, a considerably large-sized image may be hard to be reconstructed through the common CS method with some ordinary algorithms, due to the high computational complexity and hardware costs. Thus, block-based compressive sensing (BCS) is adopted as an alternative means. In our work, three scanning patterns were utilized to generate undersampled AFM images at various sampling rates for five samples with different surface appearance. Each image block was reconstructed through one of six representative algorithms. Finally, an optimal measurement scheme for BCS-HS-AFM was put forward based on the analysis of experimental results, which consisted of sampling rate, scanning pattern, block mode and reconstruction algorithm. BCS was compared with other CS methods in terms of the computational complexity and image reconstruction effect to reveal its superior performance. In brief, BCS can be applied to effectively realize the high-resolution and low-damage HS-AFM imaging.