Computational Hyperspectral Phase Retrieval for Band-Limited Spectral Imaging

Abstract

<p>We consider an application of a recently developed hyperspectral phase retrieval (HSPhR) algorithm for band-limited spectral imaging, where only six wavelengths are used instead of a continuous spectrum. The HS phase retrieval problem is considered as a reconstruction of the HS complex-valued object from encoded real-valued intensity observations with Gaussian noise. HSPhR is an iterative algorithm based on a complex domain version of the alternating direction method of multipliers, the spectral proximity operators derived for Gaussian observations, and complex-valued 3D denoising. The simulation test and experimental results demonstrate high-quality spectral resolution and phase imaging.</p>