Lune

AAAI2025Top-tier venue

Sp3ctralMamba: Physics-Driven Joint State Space Model for Hyperspectral Image Reconstruction

Ge Meng, Jingyan Tu, Jingjia Huang, Yunlong Lin, Yingying Wang, Xiaotong Tu, Yue Huang, Xinghao Ding

Abstract

Hyperspectral image (HSI) reconstruction aims to restore the original 3D HSIs from the 2D hyperspectral snapshot compressive images (SCIs). The key to high-fidelity HSI reconstruction lies in designing refined spatial and spectral attention mechanisms, which are crucial for generating finegrained representations of HSI based on the limited spatial and spectral information available in SCI. Recently, Mamba has demonstrated remarkable performance and efficiency in modeling spatial correlations. Its implicit attention mechanism generates three orders of magnitude more attention matrices than transformers, significantly raising the performance ceiling for HSI reconstruction. In this paper, we propose a novel joint SSM network named Sp3ctralMamba for HSI reconstruction. Sp3ctralMamba integrates frequency domain knowledge and physical priors to enhance reconstruction quality. Specifically, we first perform hierarchical decomposition of the 3D HSI embedding to mitigate the negative impact of distant bands on reconstruction. Next, we design a joint SSM block S 3 Mamba (S 3 MAB) to perform parallel scans of the embeddings from different bands. In addition to the conventional vanilla scan, S 3 MAB introduces a local scanning scheme to address the reconstruction challenges posed by the spatial sparsity of spectral information. Furthermore, a spiral scanning scheme in the frequency domain is incorporated to enhance the order correlation between different frequency signals. Finally, we introduce energy priors and structural priors to constrain the generation of spectral and spatial representations during the training process. Extensive experiments on both simulated and real datasets demonstrate that Sp3ctralMamba significantly elevates HSI reconstruction performance to a new level, surpassing SOTA methods in both quantitative and qualitative metrics.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

3 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model cites Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
  • Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model cites CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows
  • LRDUN: A Low-Rank Deep Unfolding Network for Efficient Spectral Compressive Imaging cites Degradation-Aware Unfolding Half-Shuffle Transformer for Spectral Compressive Imaging
  • LRDUN: A Low-Rank Deep Unfolding Network for Efficient Spectral Compressive Imaging cites Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
  • LRDUN: A Low-Rank Deep Unfolding Network for Efficient Spectral Compressive Imaging cites Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image Reconstruction
  • Degradation-Aware Unfolding Half-Shuffle Transformer for Spectral Compressive Imaging cites Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
  • Degradation-Aware Unfolding Half-Shuffle Transformer for Spectral Compressive Imaging cites Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image Reconstruction
  • CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows cites Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
  • Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image Reconstruction cites Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers3

  1. Residual Prior-driven Frequency-aware Network for Image FusionZheng Guan, Xue Wang, Wenhua Qian, Peng Liu et al.ACM MM 2025 · 10 citations
  2. SceneDecorator: Towards Scene-Oriented Story Generation with Scene Planning and Scene ConsistencyQuanjian Song, Donghao Zhou, Jingyu Lin, Fei Shen et al.NeurIPS 2025 · 9 citations
  3. LRDUN: A Low-Rank Deep Unfolding Network for Efficient Spectral Compressive ImagingHE HUANG, Yujun Guo, Wei HeCVPR 2026 · 3 citations

Builds on11

  1. Swin Transformer: Hierarchical Vision Transformer using Shifted WindowsZe Liu, Yutong Lin, Yue Cao, Han Hu et al.ICCV 2021 · 31,683 citations
  2. Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space ModelLianghui Zhu, Bencheng Liao, Qian Zhang, Xinlong Wang et al.ICML 2024 · 1,725 citations
  3. CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped WindowsXiaoyi Dong, Jianmin Bao, Dongdong Chen, Weiming Zhang et al.CVPR 2022 · 1,207 citations
  4. Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image ReconstructionYuanhao Cai, Jing Lin, Xiaowan Hu, Haoqian Wang et al.CVPR 2022 · 310 citations
  5. Degradation-Aware Unfolding Half-Shuffle Transformer for Spectral Compressive ImagingYuanhao Cai, Jing Lin, Haoqian Wang, Xin Yuan et al.NeurIPS 2022 · 222 citations

Related papers

  1. M3SR: Multi-Scale Multi-Perceptual Mamba for Efficient Spectral ReconstructionYuze Zhang, Lingjie Li, Qiuzhen Lin, Zhong Ming et al.AAAI 2026
  2. HiFi-Mamba: Dual-Stream ?-Laplacian Enhanced Mamba for High-Fidelity MRI ReconstructionHongli Chen, Pengcheng Fang, Yuxia Chen, Yingxuan Ren et al.AAAI 2026 · 2 citations
  3. VmambaSCI: Dynamic Deep Unfolding Network with Mamba for Compressive Spectral ImagingMingjin Zhang, Longyi Li, Wenxuan Shi, Jie Guo et al.ACM MM 2024 · 13 citations
  4. Detail Matters: Mamba-Inspired Joint Unfolding Network for Snapshot Spectral Compressive ImagingMengjie Qin, Yuchao Feng, Zongliang Wu, Yulun Zhang et al.AAAI 2025 · 21 citations
  5. Cycle-Consistent Mamba-Based Registration-Fusion Joint Network for Unregistered Hyperspectral Image Super-ResolutionQuangui He, Jiahui Qu, Wenqian Dong, Song Xiao et al.ACM MM 2025
Dusk over the sea between two cliffs drawn in fine vertical lines