Noise-robust phase unwrapping for smooth phase fields with physics-consistent constraints.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning for Scientific Applications

Background:

• Phase unwrapping (PU) is crucial for quantitative optical measurement and computational imaging, recovering absolute phase (AP) from wrapped phase (WP).
• Measurement noise significantly degrades the robustness of conventional PU methods, which often lack physical constraints and interpretability.
• Existing approaches struggle with noise, relying on heuristic regularization or direct mapping without physical consistency.

Purpose of the Study:

• To develop a noise-resistant phase unwrapping method that incorporates physical constraints for improved robustness and interpretability.
• To address limitations of current PU techniques in handling noisy phase data.
• To enhance the accuracy of quantitative optical measurements and computational imaging under realistic noisy conditions.

Main Methods:

• Proposed PUN-LP, a physics-consistent PU network featuring an enlarged receptive field.
• Integrated a visual state-space model for long-range spatial dependency capture and gated bottleneck convolution for local phase variation modeling.
• Introduced a hybrid loss combining multi-scale data-consistency supervision with constraints from the forward phase wrapping model.

Main Results:

• PUN-LP demonstrated improved robustness and generalization in noisy environments on synthetic and real-world datasets.
• The physics-consistent approach enforced adherence to the underlying physical laws of phase wrapping.
• Achieved superior performance compared to existing phase unwrapping methods in the presence of noise.

Conclusions:

• PUN-LP offers a robust and physically consistent solution for phase unwrapping in noisy conditions.
• The method enhances the reliability of quantitative optical measurements and computational imaging.
• Future work will focus on extending the framework to handle sharp discontinuities and complex fringe structures.