FastVSDF

Code for FastVSDF: An Efficient Spatiotemporal Data Fusion Method for Seamless Data Cube

Overview

FastVSDF is an efficient spatiotemporal data fusion method.

The purpose of proposing FastVSDF is to achieve data fusion at the lowest cost. Therefore, FastVSDF requires only minimal input data, while the computational cost needed is low.

It takes a fine/coarse image pair at T1 and coarse image at T2 to predict fine image at T2:

Get Started

Prepare Environment [Python>=3.6]

1. Download source code from GitHub.

   git clone https://github.com/ChenXuAxel/FastVSDF
   
   cd FastVSDF && git checkout release
   

2. Install dependencies.

   conda install gdal guided_filter_pytorch scikit-learn scikit-image pytorch 
   

Predict

from FastVSDF import FastVSDF
FastVSDF(F1_path, C1_path, C2_path, fastvsdf_path)

Cite

If you find FastVSDF is helpful, please cite the following work:

1. FastVSDF [Paper] [Code]

@ARTICLE{10399795,
  author={Xu, Chen and Du, Xiaoping and Fan, Xiangtao and Jian, Hongdeng and Yan, Zhenzhen and Zhu, Junjie and Wang, Robert},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={FastVSDF: An Efficient Spatiotemporal Data Fusion Method for Seamless Data Cube}, 
  year={2024},
  doi={10.1109/TGRS.2024.3353758}}

2. VSDF [Paper] [Code]

@article{XU2022113309,
title = {VSDF: A variation-based spatiotemporal data fusion method},
journal = {Remote Sensing of Environment},
volume = {283},
pages = {113309},
year = {2022},
issn = {0034-4257},
doi = {https://doi.org/10.1016/j.rse.2022.113309},
}

Contact

If you have any question, please contact Chen Xu(xuchen@aircas.ac.cn) or submit a issue.