Hybrid Deep-Neural Network in Digital Holographic Microscopy

Holography is an advanced coherent imaging technique that allows to reconstruct wavefield in both amplitude and phase, and can be applied to micro-objects using Digital Holographic Microscopy (DHM). Thanks to a digital image processing [1], aberrations can be compensated and digital refocusing can be achieved without mechani-cal displacement. Fast and simple numerical calculus can be implemented using the Angular Spectrum Method (ASM), and recently deep neural networks (DNN s) increased drastically performances of DHM techniques [2].