Deep learning-based digital in-line holographic microscopy for high resolution with extended field of view. (May 2019)

• Record Type: Journal Article
• Authors: Byeon, Hyeokjun; Go, Taesik; Lee, Sang Joon
• Is Part Of: Optics & laser technology. Volume 113(2019)
• Page Start: 77
• Page End: 86
• Publication Date: 2019-05
• Subjects: Digital in-line holographic microscopy -- 3D measurement -- Artificial intelligence; Optics -- Periodicals; Lasers -- Periodicals; Electronic journals; 621.366
• DOI: 10.1016/j.optlastec.2018.12.014
• Languages: English
• ISSNs: 0030-3992

Abstract

Highlights: Digital in-line holographic microscopy with AI-based upscaling method was proposed. AI-based method was verified by holograms from simulations and experiments. Various types of 3D information were obtained from particle/cell samples. Abstract: A digital in-line holographic microscopy (DIHM) with deep learning-based upscaling method is proposed to overcome the trade-off between high resolving power and large field-of-view (FOV). To enhance the spatial resolution of a hologram, a deep neural network was trained with hologram images, which are defocused images with diffraction patterns. The performance of the artificial intelligence-based DIHM method was verified using hologram images obtained by computer simulation and experiments. Upscaled holograms with enhanced image contrast and clear diffraction pattern provided high quality of reconstructed holograms. In addition to the enhancement of reconstructed image at the sample position, details of light scattering pattern could be revealed with the proposed method. The proposed deep learning-based DIHM method is promising for accurate monitoring of many samples and analyzing dynamics of particles or cells in large FOV with detailed 3D information reconstructed from the upscaled holograms.