High-quality Computational Ghost Imaging with Multi-Scale Light Fields Optimization

High-quality computational ghost imaging under low sampling rates has always attracted much attention and plays an important role in practical applications. In this paper, a novel optical field optimization method based on multi-scale light fields singular value decomposition which can greatly reduce the number of computational ghost imaging measurements is proposed. The computational ghost imaging measurement matrix is derived from the components obtained by singular value decomposition of self-designed special measurement matrices. When the measurement matrix is fully sampled, high-quality reconstructed image can be obtained. Similarly, when the measurement matrix is under-sampled, it is still possible to obtain high-quality reconstructed image and show the performance of multi-resolution imaging. Simulation and experimental results show that our method can obtain high-quality computational ghost imaging, even at low sampling rates, and as the number of splicing matrices increases, the number of measurements is further reduced.