HINER

This repository provides the official PyTorch implementations for the HINER series of hyperspectral image neural representations.

The code is organized into two subprojects:

Folder	Description
hiner/	Implementation of HINER: Neural Representation for Hyperspectral Image. It focuses on hyperspectral image compression and downstream classification on compressed HSI.
hiner++/	Implementation of HINER++ / Compression as Restoration. It extends implicit HSI representation to compression and restoration tasks, including denoising, inpainting, spatial super-resolution, and spectral super-resolution.

For installation, datasets, training commands, and task-specific usage, please refer to the README inside each subfolder:

• hiner/README.md
• hiner++/readme.md

Citation

If this repository is useful for your research, please cite the relevant paper.

@inproceedings{shi2024hiner,
  title={HINER: Neural Representation for Hyperspectral Image},
  author={Shi, Junqi and Jiang, Mingyi and Lu, Ming and Chen, Tong and Cao, Xun and Ma, Zhan},
  booktitle={Proceedings of the 32nd ACM International Conference on Multimedia},
  pages={9837--9846},
  year={2024}
}

@article{shi2025compression,
  title={Compression as Restoration: A Unified Implicit Approach to Self-Supervised Hyperspectral Image Representation},
  author={Shi, Junqi and Zhang, Qirui and Lu, Ming and Ma, Zhan},
  journal={IEEE Journal of Selected Topics in Signal Processing},
  year={2025},
  publisher={IEEE}
}

Acknowledgements

This codebase builds on HNeRV and SpectralFormer. We thank the authors for sharing their implementations.