Artificial Intelligence in Digital Holographic Imaging

10Noise‐free Phase Imaging in Gabor DHM with Deep Learning

10.1 Introduction

Gabor holography [1] is particularly useful in conjunction with a digital reconstruction algorithm for particle image analysis, 3D tracking, cell identification, or swimming cells in a liquid flow [2–5]. The main advantage of Gabor holography is that it can be easily set up. The optical setup is very simple and compact. In addition, its building cost is lower than those of other popular configurations in optics since it requires only a few optical components. However, Gabor holography suffers a major limitation in that a focused real image and an unfocused twin‐image are strongly superposed. To overcome this problem, several instrumental methods were proposed. However, they all required objects to stay immobile. This requirement makes it difficult to study live cells, especially in real‐time flow cytometry applications. Iterative phase‐recovery methods were also suggested for Gabor holography to remove the twin‐image noise [6–9]. Their main drawback is that they need several back‐and‐forth propagations of light to obtain the phase value. Furthermore, they also require a convergence criterion, which is generally unknown. The determination of the criterion is particularly difficult for studying biological samples in real‐time. Non‐iterative methods and inverse problem solutions were also suggested for phase recovery in Gabor holography [8–14]. Another approach in DHM is off‐axis recording, in which the ...