Random two-frame interferometry based on deep learning.

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Wavefront Sensing

Background:

• Phase-shifting interferometry (PSI) is crucial for precise wavefront measurement.
• Traditional PSI methods often require multiple frames and known phase steps.
• Accurate wavefront reconstruction from limited interferograms remains a challenge.

Purpose of the Study:

• To develop a robust two-frame phase-shifting interferometric wavefront reconstruction method using deep learning.
• To enable accurate phase calculation from two interferograms with unknown and arbitrary phase steps.
• To improve the performance and flexibility of wavefront reconstruction techniques.

Main Methods:

• A deep learning model trained on extensive simulation data derived from a physical model.
• Utilizing two interferograms with an unknown phase step for wavefront reconstruction.
• Employing a simple normalization preprocessing step, eliminating the need for DC term subtraction.

Main Results:

• The proposed deep learning method accurately reconstructs the wrapped phase from two interferograms.
• The method accommodates unknown phase steps (excluding multiples of π) and flexible interferogram sizes.
• Simulations and experiments demonstrate superior performance compared to existing two-frame methods.

Conclusions:

• The deep learning-based two-frame method provides an accurate and efficient solution for wavefront reconstruction.
• This approach simplifies the interferometric process by handling unknown phase steps and reducing preprocessing requirements.
• The method shows significant potential for advancing optical metrology and imaging applications.