High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet

To solve the problem of poor quality in ghost imaging via sparsity constraints (GISC) multispectral image reconstruction with correlation operations and compressed sensing algorithms under low sampling rate detection conditions, we propose an end-to-end deep-learning-based method. Based on the U-Net...

Full description

Saved in:
Bibliographic Details
Main Authors: Tao Hu, Jianxia Chen, Shu Wang, Jianrong Wu, Ziyan Chen, Zhifu Tian, Ruipeng Ma, Di Wu
Format: Article
Language:English
Published: IEEE 2023-01-01
Series:IEEE Photonics Journal
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10132552/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849415935246467072
author Tao Hu
Jianxia Chen
Shu Wang
Jianrong Wu
Ziyan Chen
Zhifu Tian
Ruipeng Ma
Di Wu
author_facet Tao Hu
Jianxia Chen
Shu Wang
Jianrong Wu
Ziyan Chen
Zhifu Tian
Ruipeng Ma
Di Wu
author_sort Tao Hu
collection DOAJ
description To solve the problem of poor quality in ghost imaging via sparsity constraints (GISC) multispectral image reconstruction with correlation operations and compressed sensing algorithms under low sampling rate detection conditions, we propose an end-to-end deep-learning-based method. Based on the U-Net, Res2Net-SE-Conv is employed instead of convolutional blocks to extract local and global image features at a more fine-grained level while adaptively adjusting the channel feature response. The two-dimensional contextual transformer is constructed to fully use contextual correlation information to enhance the effectiveness of feature representations. We employ the two-dimensional contextual transformer in the decoder part, dubbed CoT-Unet, to reconstruct the desired 3D cube. The results show that compared with U-Net, TSA-Net based on spatial-spectral self-attention, the PSNR of reconstructed images by the CoT-Unet is improved by 5 dB and 3 dB, respectively, SSIM is improved by 0.23 and 0.07, and SAM is decreased by 0.06 and 0.58. Compared with conventional algorithms such as DGI and CS, our method significantly improves the quality of reconstructed images. Furthermore, the comparison results at 10%, 20%, and 30% sampling rates show that our approach has the best quality in reconstructing GISC multispectral images at low sampling rates.
format Article
id doaj-art-6bac2419b3a546b4ac8c6b9d02ccf097
institution Kabale University
issn 1943-0655
language English
publishDate 2023-01-01
publisher IEEE
record_format Article
series IEEE Photonics Journal
spelling doaj-art-6bac2419b3a546b4ac8c6b9d02ccf0972025-08-20T03:33:21ZengIEEEIEEE Photonics Journal1943-06552023-01-0115311010.1109/JPHOT.2023.327938610132552High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-UnetTao Hu0https://orcid.org/0000-0003-3902-0378Jianxia Chen1https://orcid.org/0000-0001-6036-0698Shu Wang2Jianrong Wu3https://orcid.org/0000-0003-1556-7758Ziyan Chen4https://orcid.org/0000-0003-1203-7858Zhifu Tian5Ruipeng Ma6Di Wu7School of Institute of Data and Target Engineering, PLA Strategic Support Force Information Engineering University, Zhengzhou, ChinaSchool of Institute of Data and Target Engineering, PLA Strategic Support Force Information Engineering University, Zhengzhou, ChinaSchool of Institute of Data and Target Engineering, PLA Strategic Support Force Information Engineering University, Zhengzhou, ChinaKey Laboratory for Quantum Optics of CAS, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai, ChinaKey Laboratory for Quantum Optics of CAS, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai, ChinaSchool of Institute of Data and Target Engineering, PLA Strategic Support Force Information Engineering University, Zhengzhou, ChinaSchool of Cyber Science and Engineering, Zhengzhou University, Zhengzhou, ChinaSchool of Institute of Data and Target Engineering, PLA Strategic Support Force Information Engineering University, Zhengzhou, ChinaTo solve the problem of poor quality in ghost imaging via sparsity constraints (GISC) multispectral image reconstruction with correlation operations and compressed sensing algorithms under low sampling rate detection conditions, we propose an end-to-end deep-learning-based method. Based on the U-Net, Res2Net-SE-Conv is employed instead of convolutional blocks to extract local and global image features at a more fine-grained level while adaptively adjusting the channel feature response. The two-dimensional contextual transformer is constructed to fully use contextual correlation information to enhance the effectiveness of feature representations. We employ the two-dimensional contextual transformer in the decoder part, dubbed CoT-Unet, to reconstruct the desired 3D cube. The results show that compared with U-Net, TSA-Net based on spatial-spectral self-attention, the PSNR of reconstructed images by the CoT-Unet is improved by 5 dB and 3 dB, respectively, SSIM is improved by 0.23 and 0.07, and SAM is decreased by 0.06 and 0.58. Compared with conventional algorithms such as DGI and CS, our method significantly improves the quality of reconstructed images. Furthermore, the comparison results at 10%, 20%, and 30% sampling rates show that our approach has the best quality in reconstructing GISC multispectral images at low sampling rates.https://ieeexplore.ieee.org/document/10132552/Multispectral image reconstructionconvoluti onal neural networktransformerself-attention mechanismghost imaging
spellingShingle Tao Hu
Jianxia Chen
Shu Wang
Jianrong Wu
Ziyan Chen
Zhifu Tian
Ruipeng Ma
Di Wu
High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
IEEE Photonics Journal
Multispectral image reconstruction
convoluti onal neural network
transformer
self-attention mechanism
ghost imaging
title High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
title_full High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
title_fullStr High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
title_full_unstemmed High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
title_short High-Quality Multispectral Image Reconstruction for the Spectral Camera Based on Ghost Imaging via Sparsity Constraints Using CoT-Unet
title_sort high quality multispectral image reconstruction for the spectral camera based on ghost imaging via sparsity constraints using cot unet
topic Multispectral image reconstruction
convoluti onal neural network
transformer
self-attention mechanism
ghost imaging
url https://ieeexplore.ieee.org/document/10132552/
work_keys_str_mv AT taohu highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT jianxiachen highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT shuwang highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT jianrongwu highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT ziyanchen highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT zhifutian highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT ruipengma highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet
AT diwu highqualitymultispectralimagereconstructionforthespectralcamerabasedonghostimagingviasparsityconstraintsusingcotunet