Imaging through thick scattering media based on envelope-informed learning with simulated training dataset

• bin wang, Yaoyao Shi, Wei Sheng, Meiling Zhang, and Youwen Liu
• received 03/05/2024; accepted 04/24/2024; posted 04/25/2024; Doc. ID 521140
• Abstract: Computational imaging faces significant challenges in dealing with multiplescattering through thick complex media. While deep learning has addressed some ill-posedproblems in scattering imaging, its practical application is limited by the acquisition of trainingdataset. In this study, the Gaussian-distributed envelope of the speckle image is employed tosimulate the point spread function (PSF) and the training dataset is obtained by the convolutionof the handwritten digits with the PSF. This approach reduces the requirement of time andconditions for constructing the training dataset and enables a neural network trained on thisdataset to reconstruct objects obscured by an unknown scattering medium in real experiments.The quality of reconstructed objects is negatively correlated with the thickness of the scatteringmedium. Our proposed method provides a new way to apply deep learning in scattering imagingby reducing the time needed for constructing training dataset.