Deep Learning-Based Self-Interference Incoherent Digital Holography Encoding for Optical Reconstruction

We propose a self-interference incoherent holography encoding method for enhancing the optical reconstruction performance using deep learning. Self-interference incoherent digital holography records complex holograms under incoherent illumination conditions. However, conventional optical reconstruction of self-interference incoherent digital holography is conducted by phase-only modulation or amplitude-only modulation. Considering non-linear representation ability of deep neural networks, we utilize the multi-layer perceptron as a non-linear phase mapping function. We design the optimization pipeline to find the phase encoding for phase-only holographic displays and, an optical demonstration of the proposed phase encoding method is presented.