Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors

Spectral super-resolution (SSR) has garnered significant attention in recent years. Most existing networks rely on supervised methods, which require paired RGB and hyperspectral images (HSIs) for training. However, HSI acquisition is costly and time-consuming due to specialized hardware and complex...

Full description

Saved in:
Bibliographic Details
Main Authors: Xintao Zhong, Shenfu Zhang, Chenyang Lu, Xuejian Sun, Feng Shao, Weiwe Sun, Xiangchao Meng
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11098941/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849233843538624512
author Xintao Zhong
Shenfu Zhang
Chenyang Lu
Xuejian Sun
Feng Shao
Weiwe Sun
Xiangchao Meng
author_facet Xintao Zhong
Shenfu Zhang
Chenyang Lu
Xuejian Sun
Feng Shao
Weiwe Sun
Xiangchao Meng
author_sort Xintao Zhong
collection DOAJ
description Spectral super-resolution (SSR) has garnered significant attention in recent years. Most existing networks rely on supervised methods, which require paired RGB and hyperspectral images (HSIs) for training. However, HSI acquisition is costly and time-consuming due to specialized hardware and complex preprocessing. In addition, spectral mixing phenomena in low-resolution HSIs degrade image quality. To address these challenges, spectral super-resolution (SSR) techniques have emerged to generate high-quality HSIs from widely accessible RGB images, enabling applications in agriculture, medicine, and environmental monitoring. To address these issues, we propose a novel unsupervised SSR network guided by spectral sampling priors (<italic>SPointNet</italic>). Inspired by multimodality text&#x2013;image fusion techniques, we first introduce the point-image fusion module (PI-Fusion), which fuses sampled spectral data with RGB images. We then utilize spectral unmixing for super-resolution module to produce a coarse HSI, maximizing the exploitation of spectral information. Finally, we integrate a multistage shuffle-unshuffle transformer) to fuse the coarse HSI with the RGB image, enhancing its spatial information. SPointNet can ensure continuity and consistency in both spectral and spatial dimensions in the generation of the refined HSI, which is validated on three publicly available datasets.
format Article
id doaj-art-5d79e446618d4c009bfdb7a43dd9389e
institution Kabale University
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-5d79e446618d4c009bfdb7a43dd9389e2025-08-20T04:03:22ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118197921980410.1109/JSTARS.2025.359366811098941Unsupervised Spectral Super-Resolution Guided by Spectral Sampling PriorsXintao Zhong0https://orcid.org/0009-0007-6795-4932Shenfu Zhang1Chenyang Lu2https://orcid.org/0000-0002-5565-7689Xuejian Sun3Feng Shao4https://orcid.org/0000-0002-2495-9924Weiwe Sun5https://orcid.org/0000-0003-3399-7858Xiangchao Meng6https://orcid.org/0000-0002-7405-3143Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaDepartment of Geography and Spatial Information Techniques, Ningbo University, Ningbo, ChinaFaculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, ChinaSpectral super-resolution (SSR) has garnered significant attention in recent years. Most existing networks rely on supervised methods, which require paired RGB and hyperspectral images (HSIs) for training. However, HSI acquisition is costly and time-consuming due to specialized hardware and complex preprocessing. In addition, spectral mixing phenomena in low-resolution HSIs degrade image quality. To address these challenges, spectral super-resolution (SSR) techniques have emerged to generate high-quality HSIs from widely accessible RGB images, enabling applications in agriculture, medicine, and environmental monitoring. To address these issues, we propose a novel unsupervised SSR network guided by spectral sampling priors (<italic>SPointNet</italic>). Inspired by multimodality text&#x2013;image fusion techniques, we first introduce the point-image fusion module (PI-Fusion), which fuses sampled spectral data with RGB images. We then utilize spectral unmixing for super-resolution module to produce a coarse HSI, maximizing the exploitation of spectral information. Finally, we integrate a multistage shuffle-unshuffle transformer) to fuse the coarse HSI with the RGB image, enhancing its spatial information. SPointNet can ensure continuity and consistency in both spectral and spatial dimensions in the generation of the refined HSI, which is validated on three publicly available datasets.https://ieeexplore.ieee.org/document/11098941/Attention mechanismdeep learninghyperspectral image (HSI)spectral super-resolution (SSR)
spellingShingle Xintao Zhong
Shenfu Zhang
Chenyang Lu
Xuejian Sun
Feng Shao
Weiwe Sun
Xiangchao Meng
Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Attention mechanism
deep learning
hyperspectral image (HSI)
spectral super-resolution (SSR)
title Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
title_full Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
title_fullStr Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
title_full_unstemmed Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
title_short Unsupervised Spectral Super-Resolution Guided by Spectral Sampling Priors
title_sort unsupervised spectral super resolution guided by spectral sampling priors
topic Attention mechanism
deep learning
hyperspectral image (HSI)
spectral super-resolution (SSR)
url https://ieeexplore.ieee.org/document/11098941/
work_keys_str_mv AT xintaozhong unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT shenfuzhang unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT chenyanglu unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT xuejiansun unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT fengshao unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT weiwesun unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors
AT xiangchaomeng unsupervisedspectralsuperresolutionguidedbyspectralsamplingpriors