Fringe phase extraction with one step by using deep learning

Phase extraction from fringes is crucial for various optical measurement techniques. In this paper, we propose a deep learning-based phase extraction (DLPE) method to directly extract the true phase of optical fringes in a single step without the need for phase unwrapping, leveraging the advantages of deep learning in intelligent image perception. The proposed method's validity is confirmed by experiments with simulated fringes and application in extracting phase information under a real flow field using moiré deflectometry. The results clearly demonstrate that our method achieves excellent performance, characterized by high precision and structural similarity. This work eliminates the need for intermediate processing of the wrapped phase, thereby breaking the conventional two-step methods for optical fringe phase extraction. Consequently, it provides an important reference for the intelligent and efficient phase extraction from optical fringes.