Phase retrieval from experimental far-field intensity data

Abstract

We demonstrate the reconstruction of real and positive objects from experimental far-field intensity measurements by means of two phase-retrieval algorithms. Both the iterative Fourier transform and the simulated annealing algorithms are used, and an analysis is made of the advantages and disadvantages of each of these procedures and also of combinations of both methods. The objects tested either were binary or had many gray levels. We worked with data with a considerable amount of experimental noise, and in addition we recognized the importance of taking into account the nonrandom distortions produced by the detecting devices, which can critically bias the results toward erroneous estimates of the objects. Noisy data, however, can create ambiguities in the reconstructions, and hence additional information may be needed to overcome this disadvantage.