logo
CanyamAI summaries for academic research
language
MobileMobile
  • Phone FeatureWeChat Mini Program
EnglishEnglish
noticeNotice
BlogBlog
delete
arrow
Return

Construction methods of deep virtual datasets for single-plane diffractive elements

delete2025-09-16
delete0
PRE
AI
X
Xian Zhang
M
Mingxu Piao *
汪
汪俊松 (Junsong Wang)
Z
Zonglin Liang
B
Bo Zhang
DOI：10.1016/j.optlaseng.2025.109346delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
• A virtual dataset construction method was proposed for SPDOE, accurately modeling aberration- and diffraction-efficiency-induced image degradation. • A lightweight region-aware residual network was trained on the constructed dataset, achieving real-time high-quality image restoration at 10 fps. • Fabricated SPDOE with f/5, 50 mm focal length, and 486–656 nm operating band demonstrated practical validation of the proposed method. • Reconstruction achieved PSNR of 28.1 dB and SSIM of 0.9012, outperforming traditional deconvolution methods with over fivefold speedup. • The method is generalizable to SPDOEs with different optical parameters by recalculating PSF models.

Journal

Optics and Lasers in Engineering cover
Optics and Lasers in Engineering
IF:
3.7
Papers:
7.2K
Citations:
1.7W

Organization

No organization information available