Vol. 56, Issue 1, pp. 35-46 (2026)

Vol. 56 Issue 1 pp. 35-46

Unpaired self-focusing quantitative phase imaging based on cycle-GAN

Huaying Wang, Guoqing Xu, Shuo Wang, Qiaofen Zhu, Yanhui Huang

Keywords

phase imaging, deep learning, autofocus, Cycle-GAN

Abstract

In order to solve the problem that traditional microscopes cannot directly obtain the focus phase map from defocusing micrographs under incoherent illumination, a non-paired self-focusing quantitative phase imaging method based on Cycle-GAN is proposed. It only needs to obtain the bright field image of the cell by using an ordinary optical bright field microscope, and obtain the focused quantitative phase image by using the optimal weight model trained by the Cycle-GAN network. This method can directly reconstruct the phase image from an incoherent intensity map while achieving self-focusing. The quantitative analysis of the network-output phase images and holographically captured intensity images proves that the Cycle-GAN method can realize the self-focusing and quantitative phase image reconstruction of biological samples.

Vol. 56
Issue 1
pp. 35-46

1.89 MB

Corresponding address

Optica Applicata
Wrocław University of Science and Technology
Faculty of Fundamental Problems of Technology
Wybrzeże Wyspiańskiego 27
50-370 Wrocław, Poland

Publisher

Wrocław University of Science and Technology
Faculty of Fundamental Problems of Technology
Wybrzeże Wyspiańskiego 27
50-370 Wrocław, Poland

Contact us

  • optica.applicata@pwr.edu.pl
  • +48 71 320 23 93
  • +48 71 328 36 96