Automatic regularization for tomographic image reconstruction

The phase retrieval process of imaging a sample can be modeled as a simple convolution process. Sometimes, such a convolution depends on physical parameters of the sample which are difficult to estimate a priori. In this case, a blind choice for those parameters usually lead to wrong results, e.g., extracting information from the reconstructed images. In this manuscript, we propose a simple connection between phase-retrieval algorithms and optimization strategies, which lead us to ways of numerically determining the physical parameters. (C) 2019 The Author(s). Published by Elsevier B.V.