Quantitative phase imaging in digital holographic microscopy based on image inpainting using a two-stage generative
Optics Express
|October 7, 2021
Summary
This study introduces a novel phase aberration compensation method for digital holographic microscopy (DHM) using a Generative Adversarial Network (GAN). The technique accurately removes system aberrations for clearer phase imaging without complex procedures.
Area of Science:
- Optics and Photonics
- Microscopy Techniques
- Artificial Intelligence in Imaging
Background:
- Digital holographic microscopy (DHM) is a powerful imaging technique, but phase aberrations limit its precision.
- Existing methods for phase aberration compensation in DHM often involve complex procedures, prior knowledge, or manual intervention.
- Generative Adversarial Networks (GANs) offer advanced image processing capabilities with potential for novel applications in microscopy.
Purpose of the Study:
- To develop a precise and automated phase aberration compensation method for DHM.
- To leverage deep learning, specifically a two-stage GAN, for hologram inpainting and aberration removal.
- To overcome limitations of existing compensation techniques such as multiple iterations, noise sensitivity, and restricted field of view.
Main Methods:
- A two-stage Generative Adversarial Network (GAN) was employed for hologram inpainting.
- Interference fringes in the sample area were removed using edge detection and morphological image processing.
- A deep learning algorithm inpainted the removed area with generated fringes, creating a sample-free reference hologram.
- Phase aberrations were compensated by subtracting the unwrapped phase of the sample-free hologram from the inpainted results.
Main Results:
- The proposed method successfully generated a sample-free reference hologram containing system aberrations.
- Phase aberrations were effectively compensated without requiring spectrum centering, prior system knowledge, or manual input.
- The GAN-based approach demonstrated robust phase mapping, outperforming methods like Zernike polynomials fitting (ZPF) and geometrical transformations.
- Experimental validation using a silicon wafer surface measurement confirmed the method's viability and accuracy.
Conclusions:
- The developed GAN-based hologram inpainting technique provides an accurate and efficient method for phase aberration compensation in DHM.
- This approach simplifies the compensation process, making DHM more accessible and reliable.
- The study highlights the potential of GANs for advancing optical metrology and imaging applications.
