Phase retrieval of two random phase-shifting interferograms using Zernike coefficient extraction network.

Area of Science:

• Optics
• Image Processing
• Machine Learning

Background:

• Phase retrieval is crucial in optical metrology.
• Traditional methods often require complex pre-processing steps like filtering and phase unwrapping.
• Existing techniques may struggle with random phase shifts.

Purpose of the Study:

• To develop a novel deep learning method for accurate phase retrieval.
• To simplify the phase retrieval process by eliminating the need for pre-filtering and phase unwrapping.
• To address phase retrieval challenges with random phase shifts.

Main Methods:

• Phase retrieval is reformulated as a Zernike coefficient extraction problem.
• A deep learning model is employed to extract Zernike coefficients from two interferograms.
• Phase distribution is reconstructed using Zernike polynomials.

Main Results:

• The method successfully extracts Zernike coefficients from interferograms with random phase shifts.
• Simulated data analysis shows a root mean square (RMS) phase error of 0.01 λ.
• Experimental validation confirms the method's effectiveness on measured data.

Conclusions:

• The proposed deep learning method offers an efficient and accurate solution for phase retrieval.
• Elimination of pre-filtering and phase unwrapping simplifies the overall process.
• The technique demonstrates robustness and applicability to real-world optical measurements.