Virtual temporal phase-shifting phase extraction using generative adversarial networks

In this paper, a virtual temporal phase-shifting method based on generative adversarial networks (GANs) is proposed to realize dynamic measurement, which is referred to as PSNet. The proposed PSNet can produce the virtual phase-shifting fringe patterns from the single-frame fringe pattern, and the standard n-step phase-shifting method is employed to obtain the wrapped phase from the virtual fringe patterns. The wrapped phase can further be unwrapped by the unwrapping algorithm. Simulation analysis shows that the PSNet can produce the virtual phase-shifting fringe patterns without noise, and thus, the denoising process is eliminated. The performance of the method is verified from the captured fringe in the interferometer, which demonstrates the practicability of the proposed method.