HIFToolBox

Goal

Welcome to our groundbreaking project! Our mission is to compile state-of-the-art (SOTA) algorithms in the field of hyperspectral and multispectral/panchromatic sharpening. Our comprehensive collection spans three categories: model-based algorithms, unsupervised/self-supervised learning algorithms, as well as supervised learning algorithms. Join us in pioneering the advancement of spectral image processing technology!

本项目旨在收录高光谱和多光谱/全色锐化领域的SOTA算法。收录算法主要分三类：基于模型的算法、自/无监督学习算法,以及监督学习算法。加入我们，共同推动光谱图像处理技术的发展！！！！

News📰

[2026.08.10] EMR-Diff, PSRF-DiffNet, and BHSR-Net, three supervised hyperspectral image fusion methods, have been added to HIFTool with unified training and inference interfaces.

[2025.10.21] Fixed a bug that prevented HySure from supporting blind fusion.

[2025.10.15] DTDNML, BUSI, unsupervised blind methods, have been added to our collection of SOTA algorithms.

[2025.10.05] Pre-trained models on the CAVE and HARVARD datasets are released.

[2025.10.02] CaFormer, a supervised-unsupervised combined method, has been added to our collection of SOTA algorithms.

Algorithm List⚡

Model-based	Learning-based
HySure [Endnote ] HyMS [Endnote ]	supervised unsupervised MoG-DCN [Endnote ] UDALN [Endnote ] Fusformer [Endnote ] DBSR [Endnote] PSRT [Endnote ] FeafusFormer [Endnote ] MSST-Net [Endnote ] DTDNML [Endnote] DCTransformer [Endnote ] BUSI [ EndNote] EMR-Diff PSRF-DiffNet BHSR-Net ZSL UALT[Endnote ] PSTUN [Endnote ] CaFormer [Endnote ] PUT

Dataset

Chikusei Dataset
Recommended Pansharpening Datasets

How to use?

中文说明

History📰

[2025.04.11] PSTUN, a new supervised learning method, has been added to our collection of SOTA algorithms.

[2025.01.17] Compatible with CAVE and HARVARD datasets (Network_training_VL.py and Network_eval_VL.py).

[2025.01.15] UALT, a supervised-unsupervised combined method, has been added to our collection of SOTA algorithms.

[2024.05.16] Pre-trained models on the QB dataset in Pancollection Dataset are released.

[2024.05.05] Compatible with Pancollection Dataset.

[2024.04.03] DBSR, a blind unsupervised learning method, has been added to our collection of SOTA algorithms.

[2024.03.05] DCTransformer, a new supervised learning method, has been added to our collection of SOTA algorithms.

For any questions, feedback, or issues related to the HIFTool, please feel free to contact me via email.
关于HIFTool的任何问题、反馈或疑问，请随时发起issue或电子邮件与我联系。