Total Variation-Based Image Decomposition and Denoising for Microscopy Images

Area of Science:

• Microscopy
• Image Processing
• Computational Science

Background:

• Microscopy images suffer from noise, degrading quality and obscuring features.
• Increasing image acquisition rates necessitate advanced denoising and restoration techniques.
• Existing methods may not fully address the complexities of diverse microscopy data.

Purpose of the Study:

• To develop and evaluate a total variation (TV) based workflow for microscopy image decomposition and denoising.
• To assess the performance of different TV-based methods (TV-L1, Huber-ROF, TGV-L1) on various microscopy image types.
• To provide a flexible and widely applicable solution for improving microscopy image quality.

Main Methods:

• Image restoration via signal component extraction and subtraction.
• Denoising using total variation (TV) regularization.
• Comparative evaluation of TV-L1, Huber-ROF, and TGV-L1 algorithms on Atomic Force Microscopy (AFM), Scanning Tunneling Microscopy (STM), and Scanning Electron Microscopy (SEM) images.

Main Results:

• Huber-ROF demonstrated superior flexibility across different microscopy image types.
• TGV-L1 proved most effective for direct image denoising.
• The proposed TV-based workflow showed applicability beyond STM, AFM, and SEM.

Conclusions:

• The TV-based image processing workflow offers a robust solution for enhancing microscopy image quality.
• The choice between Huber-ROF and TGV-L1 depends on whether image decomposition or direct denoising is prioritized.
• Publicly available Python code (AiSurf) facilitates integration into experimental workflows for real-time or post-acquisition image enhancement.