Fig. 24: Flowchart of the phase-aberration compensation method combining Zernike polynomial fitting and deep learning, and the phase-aberration compensation results of different methods.

a The flowchart of CNN + (Zernike polynomial fitting) phase-aberration compensation. b The input phase-aberration map. c The unwrapped phase overlaid with CNN’s output, where the background (color denoted) is fed into Zernike polynomial fitting. d The 3D phase after compensation by CNN + Zernike polynomial fitting. e The phase profile of PCA²²⁶ and CNN + Zernike polynomial fitting along the dashed line in (d). a–e Adapted with permission from ref. ³⁷⁴, Optica Publishing