Background segmentation and aberration fitting via deep learning for accurate phase correction in digital holographic...

Area of Science:

• Optics and Photonics
• Computational Imaging
• Metrology

Background:

• Digital holographic interferometry is a precise measurement technique.
• Phase correction is crucial for accurate results but faces challenges with existing methods.
• Current methods struggle to balance accuracy and reliance on labeled data.

Purpose of the Study:

• To develop a novel framework for accurate phase correction in digital holographic interferometry.
• To overcome limitations of traditional and deep learning-based phase correction methods.
• To reduce dependence on labeled data while improving accuracy.

Main Methods:

• Proposed a joint neural network framework (MNN-PCNN) with a mask neural network (MNN) and phase correction neural network (PCNN).
• MNN performs background segmentation to isolate object interference.
• PCNN utilizes Zernike polynomial modes and self-supervised learning for aberration fitting and noise suppression.

Main Results:

• MNN-PCNN demonstrated superior performance compared to classical polynomial fitting and UNet.
• The framework effectively corrects phase deviations in digital holographic micro-tomography.
• Achieved enhanced accuracy in quantitative three-dimensional measurements.

Conclusions:

• The MNN-PCNN framework offers an effective solution for phase correction in digital holography.
• It provides a balance between accuracy improvement and reduced reliance on labeled data.
• Significantly enhances the precision of quantitative 3D measurements in micro-tomography.