Reduction of artifacts associated with missing data in coherent diffractive imaging

Abstract

Coherent diffractive imaging experiments often collect incomplete datasets containing regions that lack any measurements. These regions can arise because of beamstops, gaps between detectors, or, in tomography experiments, a missing wedge of data due to a limited sample rotation range. We describe practical and effective approaches to mitigate reconstruction artifacts by bringing uniqueness back to the phase retrieval problem. This is accomplished by looking for a solution that both matches the data and has minimum total variation, which essentially sets the unconstrained modes to reduce oscillations within the reconstruction. Two algorithms are described. The first algorithm assumes that there is an accurate estimate of the phase and can be used for pre- and post-processing. The second algorithm attempts to simultaneously minimize the total variation and recover the phase. We demonstrate the utility of these algorithms with numerical simulations and, experimentally, on a large, three-dimensional dataset.