Computational ghost imaging with compressed sensing based on a convolutional neural network

Computational ghost imaging (CGI) has recently been intensively studied as an indirect technique. However, the speed of CGI cannot meet requirements practical applications. Here, we propose a novel scheme for high-speed imaging. In our scenario, conventional data processing algorithm is optimized to new compressed sensing (CS) based on convolutional neural network (CNN). CS used process collected by device. Then, processed are trained CNN reconstruct image. The experimental results show that can produce high-quality images with much less sampling than CGI. Moreover, detailed comparisons between reconstructed using approach and deep learning (DL) outperforms achieves faster speed.