RS-UNet++-based phase distortion compensation in digital holography

Area of Science:

• Optics and Photonics
• Image Processing
• Machine Learning

Background:

• Digital holographic imaging provides 3D information but is susceptible to phase distortions.
• Misalignment, environmental factors, and component differences introduce errors affecting measurement accuracy.

Purpose of the Study:

• To develop an accurate phase distortion compensation method for digital holographic measurements.
• To improve the precision of 3D information recovery in holographic imaging.

Main Methods:

• Proposed a Residual Squeeze-and-Excitation Nested U-Net Architecture (RS-UNet++).
• Utilized histogram threshold segmentation and random library selection for phase distortion data extraction.
• Trained the network using combined phase data and generated distortion data.

Main Results:

• RS-UNet++ demonstrated superior phase distortion compensation compared to IRLS, Poly-fit, PCA, CNN, and Res-UNet.
• Achieved optimal results in standard deviation, global error, and Pearson's correlation coefficient.
• Significantly improved the accuracy of phase recovery in simulations and experiments.

Conclusions:

• The proposed RS-UNet++ method effectively compensates for phase distortions in digital holographic imaging.
• This advancement leads to substantially more accurate 3D measurements.
• The method offers a robust solution for enhancing holographic measurement precision.