Optimization-techniques for efficient measurement of the transmission-matrix of scattering media
Master of Science
Scattering-media occur in a wide-variety of imaging applications: atmospheric turbulence in astronomical imaging, turbid-tissues in medical imaging, fog in navigational imaging, etc. In this thesis we investigate imaging through scattering media using transmission-matrices (TMs); complex-valued matrices that characterize the input output relationship of an electromagnetic wavefront as it passes through a multiply scattering media. Computing the TM of a scattering-media requires measuring the amplitude and phase-information of a series of light-waves used for calibration. Traditionally these wavefronts are measured using hard-to-setup interferometric systems. We use the much easier to setup double phase-retrieval technique to get access to the same. This method reconstructs complex-valued fields from real-valued measurements using phase retrieval algorithms. Phase retrieval algorithms can be very sensitive to the type of measurement matrix used: Oftentimes algorithms are derived assuming i.i.d.~Gaussian measurements and when these conditions are not satisfied performance suffers. In this thesis we circumvent this issue by developing a new algorithm, prVAMP, that demonstrates state-of-the art performance with realistic measurements. Using prVAMP, we construct high resolution TMs for ground-glass diffusers. We then investigate several properties of these TMs and discover a new one: TMs can be decomposed into a series of speckle-patterns which shift at different rates. In simulation, we demonstrate that this property has significant implications.
Phase retrieval; transmission-matrix; speckle-patterns; Gaussian-measurement