Skip to content

Latest commit

 

History

History
73 lines (49 loc) · 2.82 KB

README.md

File metadata and controls

73 lines (49 loc) · 2.82 KB

Laplacian pyramid networks: A new approach for multispectral pansharpening

Homepage:

https://academic.peterkam.top/

https://liangjiandeng.github.io/

https://sites.google.com/site/vivonegemine/

  • Code for paper: "Laplacian pyramid networks: A new approach for multispectral pansharpening, Information Fusion"
  • State-of-the-art pansharpening performance

Dependencies and Installation

  • Python 3.8 (Recommend to use Anaconda)
  • TensorFlow 1.14.0
  • NVIDIA GPU + CUDA
  • Python packages: pip install numpy scipy h5py
  • TensorBoard

Dataset Preparation

The datasets used in this paper is WorldView-3 (can be downloaded here), QuickBird (can be downloaded here) and GaoFen-2 (can be downloaded here). Due to the copyright of dataset, we can not upload the datasets, you may download the data and simulate them according to the paper.

Get Started

Training and testing codes are in 'codes/'. Pretrained model can be found in 'codes/pretrained/'. All codes will be presented after the paper is completed published. Please refer to codes/how-to-run.md for detail description.

LPPN Architecture

LPPN_architecture

FCNN architecture is presented below:

FCNN_architecture

Results

Quantitative results

The following quantitative results is generated from WorldView-3 datasets. A.T. is short for Average running Time for saving spaces in the paper.

Quantitative_WV3

All quantitative results can be found in 'results/'.

Visual Results

The following visual results is generated from WorldView-3 datasets.

Visual_WV3

All visual results can be also found in 'results/'.

Acknowledgement

Part of code of this work is derived from https://xueyangfu.github.io/projects/LPNet.html.

Citation

@article{LPPN,
author = {Cheng Jin, Liang-Jian Deng, Ting-Zhu Huang and Gemine Vivone},
title = {Laplacian pyramid networks: A new approach for multispectral pansharpening},
journal = {Information Fusion},
volume = {78},
pages = {158-170},
year = {2022},
issn = {1566-2535},
doi = {https://doi.org/10.1016/j.inffus.2021.09.002}
}

Contact

We are glad to hear from you. If you have any questions, please feel free to contact Cheng.Jin@std.uestc.edu.cn or open issues on this repository.

License

This project is open sourced under GNU Affero General Public License v3.0.