Improving 3D Recovery based on Super-Resolution Generative Adversarial Network and Uniform Continuous Trajectory for Atomic Force Microscopy

Atomic force microscope (AFM) is a powerful nano-scale measurement instrument, which is diffusely applied on different fields, such as biological science, nanomanipulation, semiconductor, Micro Electro Mechanical Systems (MEMS) detection, etc. The well-known advantage of AFM is its high-accuracy 3D topography reconstruction. Different from optical microscopy, which can only obtain 2D image by optical reflection, three kinds of operating principles of AFM respectively maintaining the contact force, amplitude or distance between the tip and sample surface during scanning to collect the sample's height information, and then help us to build a 3D sample topography. However, because of the physical contact with probe, there is a major problem in AFM - imaging speed. In this paper, we propose a new method which applies the Generative Adversarial Networks (GAN) to AFM image reconstruction, which can recover a high-resolution (HR) image from a low-resolution (LR) one with only a quarter of time. While using GAN, data uniformity is most crucial. To address this issue, we propose a new trajectory - Uniform continuous path (UC path) to break the limits on traditional raster scanning and a proposed feature similarity metric is used on comparing the reconstruction results in experiments.