Analysis of phase retrieval from multiple images

Abstract

This thesis considers the calculation of phase from sets of phase contrast and defocused images. An improvement to phase contrast imaging is developed that combines three phase contrast images. This method results in a reduction in the phase error by a factor of up to 20 in comparison to using a single image. Additionally the method offers the potential for optimisation and the extension to an arbitrary number of images. Phase diversity using defocused images is considered in more depth where the intensity transport equation is used to calculate the phase. First a Green's function approach to solving this equation was considered. One of the Green's functions stated in the literature is shown to be incorrect, the other two are shown to be correct both giving equivalent phase estimates. A further improvement is made to this method by removing the singularities in the phase calculation process. As an alternative to the Green's function solution a Fourier transform approach is also considered. A complete solution to the intensity transport equation is derived with inclusion of the boundary conditions. This completes the method incompletely described in the literature. Through simulation, generic key factors are identified for the potential optimisation of experimental and numerical process to improve the estimated phase. Determining 3D structural information of an object from the phase calculated in a single plane is considered using an iterative process. It is shown that this process is limited but can be used, in some cases, to generate an approximate representation of the object.