Phase retrieval method for single-frame point diffraction interferogram images based on deep learning.

Area of Science:

• Optical Metrology
• Artificial Intelligence in Optics
• Interferometry

Background:

• Traditional point diffraction interferometry suffers from reduced accuracy due to environmental errors from multi-step phase shifting.
• Accurate phase retrieval is crucial for high-precision optical metrology and wavefront sensing.

Purpose of the Study:

• To develop a deep-learning-based phase retrieval method for single-frame point diffraction interferograms.
• To enhance measurement accuracy and overcome limitations of traditional interferometric techniques.
• To provide a fast and accurate solution for phase retrieval in point diffraction interferometry.

Main Methods:

• Proposed a novel phase retrieval method utilizing deep learning for single-frame point diffraction interferograms.
• Constructed two specialized neural networks for interference fringe image processing.
• Developed a diverse dataset of point diffraction images for training and optimization.

Main Results:

• The deep-learning method achieved accurate and rapid phase retrieval, enabling high-precision phase unwrapping.
• Validation using experimental data showed results consistent with professional interferogram processing software (ESDI) and other algorithms.
• Demonstrated significant improvement in measurement accuracy compared to traditional methods.

Conclusions:

• The proposed deep-learning approach offers a feasible and effective solution for high-precision phase retrieval in point diffraction interferometry.
• This method addresses environmental error issues inherent in multi-step phase shifting techniques.
• The approach is both fast and highly accurate, suitable for advanced optical metrology applications.