Phase Retrieval with Sparsity Constraints

by Stefan Loock
Doctoral thesis
Date of Examination:2016-06-07
Date of issue:2016-06-29
Advisor:Prof. Dr. Gerlind Plonka-Hoch
Referee:Prof. Dr. Gerlind Plonka-Hoch
Referee:Prof. Dr. Russell Luke
Persistent Address: http://dx.doi.org/10.53846/goediss-5697

 

 

Files in this item

Name:Dissertation_PRwSC_Loock.pdf
Size:3.53Mb
Format:PDF

The following license files are associated with this item:

Abstract

English

The two-dimensional phase retrieval problem arises in many areas of experimental physics, e.g. in x-ray microscopy.  The central theme of this thesis is the application of sparsity constraints in the two-dimensional discrete phase retrieval problem. It provides a framework for the utilization of sparsifying transforms, such as the discrete shearlet transform, which is an extension of the wavelet transform that is especially suited for the efficient representation of so-called cartoon like images. Based on the relaxed averaged alternating reflections (RAAR) algorithm, a reconstruction algorithm  is proposed which incorporates shrinkage mappings of frame coefficients. For tight frames we show that the resulting operator is the proximity operator of a proper, lower-semicontinuous, convex function. Furthermore, bounds on the iterates of the newly developed algorithm as well as Césaro convergence are proven for arbitrary frames. The thesis concludes with a numerical evaluation of simulated measurement data for x-ray microscopy experiments in the near-field regime contaminated by Poisson noise.
Keywords: phase retrieval; sparsity constraints; projection algorithms; relaxed averaged alternating reflections; shearlets; wavelets; optimization; numerical analysis
 
Statistik