Adaptive AFM image reconstruction through frequency coefficient selection based on block compressed sensing

Abstract

Atomic force microscope (AFM) is a valuable instrument for nano-scale imaging. Traditional AFM demands long time to obtain AFM images due to point-by-point imaging. AFM based on compressed sensing (CS-AFM) is able to obtain AFM images by reconstructing from incomplete measurement. Block compressed sensing (BCS) further shortens the reconstruction time but focuses on the topography of samples, which loses necessary frequency information. Different subblocks exhibit various characteristics by sparse transformation. In this paper an adaptive frequency coefficient selection method based on BCS is proposed to enhance reconstruction quality. We focus on the frequency domain of the AFM image, applying sparse transformation to obtain the corresponding frequency coefficients. We then select partial coefficients from all frequency components as the feature information for each subblock during reconstruction, followed by an inverse sparse transformation to obtain the final reconstructed image. The proposed method outperforms both iterative selection-based BCS and ordinary BCS, achieving the highest PSNR and SSIM values across various sampling ratio while maintaining comparable reconstruction times, which verifies its ability to fast reconstruct AFM images with high quality.