Second Harmonic Imaging Enhanced by Deep Learning Decipher

Wavefront sensing and reconstruction are widely used for adaptive optics, aberration correction, high-resolution optical phase imaging. Traditionally, interference and/or microlens arrays to convert the into intensity variation. Direct imaging of distorted wavefront usually results in complicated retrieval with low contrast sensitivity. Here, a novel nonlinear encoding approach has been developed experimentally demonstrated using second harmonic generation sharpen information carried by probe beam. By designing implementing deep neural network, we demonstrate enhanced learning decipher (SHIELD) efficient resilient retrieval. Inheriting advantages two-photon microscopy, SHIELD demonstrates single-shot, reference-free, video-rate sensitivity better than λ100 high robustness against noise, facilitating numerous applications from biological sensing.