Ghost Imaging with Deep Learning for Position Mapping of Weakly Scattered Light Source

Abstract We propose ghost imaging (GI) with deep learning to improve detection speed. GI, which uses an illumination light random patterns and a single-pixel detector, is correlation-based thus suitable for detecting weak light. However, its time too long practical inspection. To overcome this problem, we applied convolutional neural network that was constructed based on classification of the causes image degradation. A feasibility experiment showed when using digital mirror device projector photodiode, proposed method improved quality images.