Single-shot and twin-image free unique phase retrieval using an aspect of autocorrelation that considers the object asymmetry.

A phase-retrieval algorithm can reconstruct the phase of an object from an intensity image of a diffraction pattern recorded at the Fourier plane, provided one makes a right initial guess. For the algorithms that are based on an initial random guess, at times, the output either becomes nondeterministic, becomes nonconvergent, or develops a twin image. For improving the performance of the algorithms by eliminating the twin image and for bringing uniqueness to the output, prior information about the object or more intensity measurements are needed. In order to achieve this using only a single acquisition of the intensity pattern recorded at the Fourier plane for an object, we propose a method in which object support, along with the object information in the case in which the object has a direction of asymmetry, can be measured from the object autocorrelation and utilized as an initial guess in the phase-retrieval algorithm. In the process, we explore how even a partial asymmetry in the object distribution can lead to a good solution in the phase-retrieval algorithm. This method is beneficial for unique phase detection without any twin artifact in various fields of optics. We theoretically analyze the proposed method and validate it by carrying out the simulations and experiment.