Image segmentation of phase-modulated holographic data storage based on deep learning

Abstract

Phase retrieval based on data-driven deep learning (DL) is a suitable decoding method for phase-modulated holographic data storage (HDS). Once the DL network is trained, the phase can be directly retrieved from the corresponding diffraction intensity image with high data transfer rate and low bit error rate. Traditional data-driven DL-based phase retrieval requires a large number of known samples for training, which is usually laborious for practical applications such as HDS. In the paper, we propose an image segmentation method based on image features, leading to about 54 times reduction in the number of original sample pairs (OSP) for training DL network. The proposed method is easy to implement in practical situations of HDS.