Algorithms for phase retrieval with a (rough) phase estimate available: a comparison

In this work we consider a variation of the phase retrieval problem where the phase is not lost completely but is known to lie within a certain interval. For this problem, we develop two new algorithms and evaluate their performance compared with the classical Hybrid Input-Output (HIO) method. One of the algorithms is a modification of the Hybrid Input-Output method which uses the additional information about the phase, and the second is based on convex optimization techniques. Extensive tests demonstrate that the latter algorithm outperforms significantly both the original HIO and its phase-aware modification. c © 2010 Optical Society of America