Removal of algorithmic stagnation by augmented iterative phase retrieval

Retrieving the phase of an optical field using intensity measurements is one of the most widespread and studied inverse problems in classical optics. However, common iterative approaches such as the Gerchberg-Saxton algorithm and its derivatives suffer from the twin-image problem – the iterative minimisation stagnates and the recovered field contains features from both the target field and its point-reflection. We present a technique that leverages mathematical properties of the stagnated field, to constrain the problem and remove the twin image artefacts. This improvement in reconstruction robustness has implications in a range of fields, including applications in adaptive optics, holography and optical communications.