Skip to content

Comparative Study of Deep Learning Algorithms for Atomic Force Microscope Image Denoising

Notifications You must be signed in to change notification settings

hoichanjung/AFM_Image_Denoising

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

71 Commits
 
 
 
 
 
 

Repository files navigation

Atomic Force MicroScope(AFM) Image Denoising

Requirements

  • Python 3.7.0
  • Cuda 11.1
  • pip install -r requirements.txt

Data Description

Filtered Image (by Seung Jun JUNG)

image

Sample Image (652 AFM Images with Random/Line/Scar/Hum Noise)

image

Paper

  • Comparative Study of Deep Learning Algorithms for Atomic Force Microscope Image Denoising (Hoichan Jung, Giwoong Han, Seong Jun Jung and Sung Won Han)

Atomic force microscopy (AFM) enables direct visualisation of surface topography at the nanoscale. However, post-processing is generally required to obtain accurate, precise, and reliable AFM images owing to the presence of image artefacts. In this study, we compared and analysed state-of-the-art deep learning models, namely MPRNet, HINet, Uformer, and Restormer, with respect to denoising of AFM images containing four types of noise. Specifically, the denoising performance and inference time of these algorithms on AFM images were compared. Peak signal-to-noise ratio and structural similarity index map were used to evaluate the denoising performance. Based on comparative analysis, Restormer is the most effective and HINet is the most efficient model for denoising AFM images.

Quantitative Result

Comparing PSNR and SSIM of Image Denoising Deep Learning Models on AFM Images with Random, Line, Scar, and Hum Noise.

image

PSNR vs. inference time on the AFM image

image

Qualitative Result

Random Noise

image

Line Noise

image

Scar Noise

image

Hum Noise

image

About

Comparative Study of Deep Learning Algorithms for Atomic Force Microscope Image Denoising

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published