Computational holographic imaging through scattering media using deep learning

It is difficult to restore the phase-only computer-generated hologram (CGH) through scattering medium since wavefront of optical beam is distorted by the speckles caused by random medium. In this article we presented Generative Adversarial architecture (GAN) used to learn functional relation between phase map imaging and ground truth imaging and then reconstruct CGH through scattering medium. Our experiments proved that GAN has learned the scattering characteristics of the scattering medium and used to restore CGH through scattering medium.