Super-resolution reconstruction of the holographically reconstructed image based on improved ESRGAN

Chen, Y, Li, S, Ma, H, Li, P, Zhang, G ORCID: 0000-0002-2351-2661, Guo, J and Dong, M (2025) Super-resolution reconstruction of the holographically reconstructed image based on improved ESRGAN. Optics Communications, 596. ISSN 0030-4018

Text manuscript-R1(Text unmarked).pdf - Accepted Version Access Restricted until 10 September 2026. Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (1MB)

Abstract

In digital holographic measurement, various noise sources, including dark current noise, readout noise, and photon shot noise from the CCD, can adversely affect the recording of interference fringes by introducing voltage fluctuations and degrading fringe contrast. Furthermore, optical system imperfections such as stray light, component defects and air turbulence lead to superposition and distortion of the fringe pattern, causing spatial deformation and temporal jitter. These combined effects ultimately compromise the resolution of the reconstructed holographic image. Therefore, in this paper, a super-resolution reconstruction method for the holographically reconstructed images based on improved ESRGAN is proposed, by optimizing the network structure and loss function, and introducing a noise reduction module to enhance the denoising ability and detail recovery performance of this method. Firstly, the DIV2K public dataset is converted into a digital holographically simulated dataset by means of the off-axis Fresnel holographic recording method. Then, the experimental dataset is constructed by reconstructing the holograms of the samples collected using the reflective off-axis optical path system based on Fresnel diffraction method. Finally, the constructed dataset is used for network training and testing, and the experimental results are compared and analyzed with those of the other four methods. The experimental results show that the proposed method outperforms the other methods in five metrics: PSNR, SSIM, RMSE, LPIPS and BRISQUE, especially in detail recovery and texture enhancement, and the reconstruction results are more in line with human visual perception characteristics. Therefore, the improved ESRGAN method proposed in this paper has significant advantages in the super-resolution reconstruction quality of the holographically reconstructed images, and provides a new technical approach for high-precision holographic imaging.

Item Type:	Article
Uncontrolled Keywords:	4006 Communications Engineering; 40 Engineering; 51 Physical Sciences; Bioengineering; 0205 Optical Physics; 0906 Electrical and Electronic Engineering; 1005 Communications Technologies; Optoelectronics & Photonics; 4006 Communications engineering; 5102 Atomic, molecular and optical physics; 5108 Quantum physics
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Divisions:	Engineering
Publisher:	Elsevier
Date of acceptance:	9 September 2025
Date Deposited:	26 Sep 2025 10:31
Last Modified:	26 Sep 2025 10:45
DOI or ID number:	10.1016/j.optcom.2025.132451
URI:	https://researchonline.ljmu.ac.uk/id/eprint/27207

View Item