Lensless Single-shot Pixel Super-resolution Phase Microscopy

Abstract

This doctoral thesis is dedicated to the development of a mini phase microscope capable of subpixel resolution without using lenses. Instead of lenses, a custom-made diffractive element, a small-pixeled binary phase mask is used along with state-of-the-art phase retrieval-based algorithms to achieve high-resolution reconstruction of complex-valued objects.

In phase retrieval, the test object is illuminated by coherent light source, which changes depending on the object’s characteristics. From this diffracted intensity pattern captured by the sensor a complex wavefront is reconstructable. Although, it is an ill-posed problem, since the traditional sensors can capture only the intensity of the light radiation, while the phase is in invisible range. Moreover, the scheme is in-line therefore, the pattern will result in several overlapping diffraction orders. The overlapping images and the ill-posedness make the reconstruction from the diffraction patterns challenging. Most methods provided by the literature try to solve this problem by registering several decorrelated diffraction patterns on the sensor – with a so-called multi-exposure method.

Contrary to those techniques, the first contribution of the thesis is the development of a single-exposure approach for phase retrieval using lensless wavefront modulation. The modulation is achieved by a single random binary phase mask positioned between the object and the sensor. The second contribution of the thesis is an optical system design for the proposed algorithm to achieve super-resolution reconstruction. The system design includes the investigation of the system parameters, like distance tuning, noise-influence analysis, and modulation mask selection. The third contribution of the thesis is a novel approach based on wavefront separation to further reduce the appearing noises and enhance the resolving power. We computationally separated the carrying and objects’ wavefronts, which was not considered before. The performance of the novel approach and algorithm is demonstrated in simulations and physical experiments. We report an experimental computational super-resolution of 2um lines of USAF phase target, which is 3.45x smaller than the resolution following from the Nyquist-Shannon sampling theorem for used camera pixel size 3.45um. To the best of our knowledge, the 2um resolution is beyond the state-of-the-art resolution for single-shot phase retrieval techniques reported so far, even with lens configurations.