Tile-Based Two-Dimensional Phase Unwrapping for Digital Holography Using a Modular Framework

Automated phase unwrapping in digital holography with deep learning

Digital holography can provide quantitative phase images related to the morphology and content of biological samples. After the numerical image reconstruction, the phase values are limited between -π and π; thus, discontinuity may occur due to the modulo 2π operation. We propose a new deep learning model that can automatically reconstruct unwrapped focused-phase images by combining digital holography and a Pix2Pix generative adversarial network (GAN) for image-to-image translation. Compared with numerical phase unwrapping methods, the proposed GAN model overcomes the difficulty of accurate phase unwrapping due to abrupt phase changes and can perform phase unwrapping at a twice faster rate. We show that the proposed model can generalize well to different types of cell images and has high performance compared to recent U-net models. The proposed method can be useful in observing the morphology and movement of biological cells in real-time applications.

Smart filtering of phase residues in noisy wrapped holograms

Phase unwrapping is one of the major challenges in multiple branches of science that extract three-dimensional information of objects from wrapped signals. In several applications, it is important to extract the unwrapped information with minimal signal resolution degradation. However, most of the denoising techniques for unwrapping are designed to operate on the entire phase map to remove a limited number of phase residues, and therefore they significantly degrade critical information contained in the image. In this paper, we present a novel, smart, and automatic filtering technique for locally minimizing the number of phase residues in noisy wrapped holograms, based on the phasor average filtering (PAF) of patches around each residue point. Both patch sizes and PAF filters are increased in an iterative algorithm to minimize the number of residues and locally restrict the artifacts caused by filtering to the pixels around the residue pixels. Then, the improved wrapped phase can be unwrapped using a simple phase unwrapping technique. The feasibility of our method is confirmed by filtering, unwrapping, and enhancing the quality of a noisy hologram of neurons; the intensity distribution of the spatial frequencies demonstrates a 40-fold improvement, with respect to previous techniques, in preserving the higher frequencies.

Compact single-shot four-wavelength quantitative phase microscopy with polarization- and frequency-division demultiplexing

We present a novel single-shot four-wavelength quantitative phase microscopy (FW-QPM). Four lasers operating at different wavelengths are multiplexed with a pair of dichroic mirrors and a polarization beam splitter in a three-mirror quasi-common-path interferometer. After a single-shot interference pattern is obtained with a monochrome camera, four holograms of different wavelengths were demultiplexed from it in the frequency domain with polarization- and frequency-division multiplexing. Polarization-division demultiplexing scheme uses polarization dependent visibility changes in an interference pattern, and it plays a critical role in making only two interference patterns exist within a single quadrant in the frequency domain. We have used a single-mode optical fiber as a phase object sample and demonstrated that a measured single-shot interference pattern can be successfully demultiplexed into four different interferograms of different wavelengths with our proposed scheme.