Depth invariant feature extraction using deep learning in strong scattering

We present that the trained phase reconstruction convolutional neural network (PRCNN) is able to extract the depth invariant information and reconstruct high-quality phase retrieval images of sparsity objects through a strong diffuser.