Fig. 8: Performance evaluation of vision aberration correction.
From: End-to-end learning of 3D phase-only holograms for holographic display
a Top: A hologram calculated for normal vision. Middle: A hologram calculated using the proposed variant of SM-LBM that corrects a synthetic astigmatic vision induced by a 200 mm cylindrical lens. The calibration method is detailed in Methods. Bottom: The CNN trained on the astigmatism-corrected dataset jointly performs diffraction simulation and aberration correction. The PSNR and SSIM of the predicted hologram are marked on the top right corner. b Experimentally captured CNN-predicted holograms of a natural scene and a test scene without (Left) and with (Right) astigmatism correction. The yellow and the blue box marks the foreground and the background object of interest, respectively. The in-focus object in the corrected hologram exhibits horizontal blurs in the uncorrected hologram (i.e., treefrog’s tentacles, eye reflections, tree leaves, and test patterns)
