Optimal sampling and reconstruction of undersampled atomic force microscope images using compressive sensing

Abstract

Atomic force microscope (AFM) is an analytical instrument which is used to study the surface structure and morphology of materials. The AFM can measure and observe samples either in air or liquid environment. However, the standard AFM requires a long time to acquire accurate images and data. In our work, the compressive sensing (CS) was applied in order to reduce the imaging time, lower the interactions between the probe and the sample, finally avoid sample damage in AFM. Three samples (PAA film, TGG1 grating and BOPP film) were used as the testing samples. Different image reconstruction algorithms (l1-ls, TVAL3, GPSR and IHT) were employed to reconstruct AFM image with different sampling rate. And various sampling patterns (Random Scan, Row Scan, SRM, Spiral Scan and Square-shape Scan) were used to obtain the undersampling data. A large number of experiments show that the choice of sampling pattern and image reconstruction algorithm has significant impact on the quality of the reconstructed images in AFM. Subsequently the reconstruction results of sample topographic images were analyzed and evaluated by the image quality indicators (PSNR and SSIM). The CS method can be used to obtain accurate images by reducing measurement data. It finally improves the measurement speed of AFM without cutting down the quality of AFM image.