基于深度学习的超光谱图像重建与超瑞利斑纹
天体物理仪器与方法
2025-02-27 v1
摘要
基于稀疏约束的幽灵成像（GISC）光谱相机将三维（3D）超光谱图像调制为二维（2D）压缩图像，形成单次快门中的斑纹。通过重建算法获得3D超光谱图像（HSI）。深度学习的快速发展为3D HSI重建提供了新方法。此外，GISC光谱相机的成像性能可通过优化斑纹调制来提升。本文提出一种采用超瑞利斑纹调制的端到端GISCnet，以提高GISC光谱相机的图像重建质量。GISCnet的结构简单而有效，我们可以轻松调整网络结构参数以提高图像重建质量。相对于瑞利斑纹，我们的超瑞利斑纹调制在重建3D HSI时展现出丰富的细节。经过对648个3D HSI的评估，发现平均峰信噪比从27 dB提升至31 dB。总体而言，本文提出的采用超瑞利斑纹调制的GISCnet能够通过优化超瑞利调制和深度学习图像重建两方面，有效提高GISC光谱相机的图像质量，为联合优化光场调制和图像重建以提升幽灵成像性能提供了灵感。
引用
@article{arxiv.2502.18776,
title = {First frequency phase transfer from the 3 mm to the 1 mm band on an Earth-sized baseline},
author = {Sara Issaoun and Dominic W. Pesce and María J. Rioja and Richard Dodson and Lindy Blackburn and Garrett K. Keating and Sheperd S. Doeleman and Bong Won Sohn and Wu Jiang and Dan Hoak and Wei Yu and Pablo Torne and Ramprasad Rao and Remo P. J. Tilanus and Iván Martí-Vidal and Taehyun Jung and Garret Fitzpatrick and Miguel Sánchez-Portal and Salvador Sánchez and Jonathan Weintroub and Mark Gurwell and Carsten Kramer and Carlos Durán and David John and Juan L. Santaren and Derek Kubo and Chih-Chiang Han and Helge Rottmann and Jason SooHoo and Vincent L. Fish and Guang-Yao Zhao and Juan Carlos Algaba and Ru-Sen Lu and Ilje Cho and Satoki Matsushita and Karl-Friedrich Schuster},
journal= {arXiv preprint arXiv:2502.18776},
year = {2025}
}
备注
11 pages, 5 figures, accepted to AJ