Phase retrieval of coherent diffractive images with global optimization algorithms

Abstract

Coherent diffractive imaging (CDI) or lensless microscopy has recently been of great interest as a promising alternative to electron microscopy in achieving atomic spatial resolution. Reconstruction of images in real space from a single experimental diffraction pattern in CDI is based on applying iterative phase-retrieval (IPR) algorithms, such as the hybrid input–output and the error reduction algorithms. For noisy data, these algorithms might suffer from stagnation or trapping in local minima. Generally, the different local minima have many common as well as complementary features and might provide useful information for an improved estimate of the object. Therefore, a linear combination of a number of chosen minima, termed a basis set, gives an educated initial estimate, which might accelerate the search for the global solution. In this study, a genetic algorithm (GA) is combined with an IPR algorithm to tackle the stagnation and trapping in phase-retrieval problems. The combined GA–IPR has been employed to reconstruct an irregularly shaped hole and has proven to be reliable and robust. With the concept of basis set, it is strongly believed that many effective local and global optimization frameworks can be combined in a similar manner to solve the phase problem.