Zhiying Jiang, Zengxi Zhang, Jinyuan Liu, Xin Fan, Risheng Liu* ,“Multi-Spectral Image Stitching via Global-Aware Quadrature Pyramid Regression”, IEEE Transactions on Image Processing (TIP), 2024. [paper]
[2024-08-01] The testing code is available
[2024-10-26] The training code in pytorch version is available
Download the pre-trained model and put it in 1_ImageAlignment_Pytorch/snapshot
Download the pre-trained model and put it in 2_ImageReconstruction_Pytorch/snapshot
(optional) Download the pre-trained model and put it in 2_Reconstruction_Tensorflow/snapshot
Download the pre-trained model and put it in 3_ImageFusion_Pytorch/snapshot
Pytorch==1.12.1 Tensorflow==2.10.1
cd 1_ImageAlignment_Pytorch
python test.py
The generated images are used to the subsequent reconstruction.
Tips: The original model was trained on tensorflow. For better use, we reproduced the whole model in Pytorch, but the image reconstruction effect deviated from the source code. Therefore, in addition to the pytorch version, we also open-sourced the source code for image reconstruction in Tensorflow. If you don't have tensorflow installed, you can also employ the Pytorch version in the experimental comparison.
cd ../2_ImageReconstruction_Pytorch
python test.py
cd ../2_ImageReconstruction_Tensorflow
python test.py
cd ../3_ImageFusion_Pytorch
python test.py
cd 1_ImageAlignment_Pytorch
python train_H.py
cd ../2_ImageReconstruction_Pytorch
python train.py
cd ../3_ImageFusion_Pytorch
python train.py
@ARTICLE{2024jianggqpr,
author={Jiang, Zhiying and Zhang, Zengxi and Liu, Jinyuan and Fan, Xin and Liu, Risheng},
journal={IEEE Transactions on Image Processing},
title={Multispectral Image Stitching via Global-Aware Quadrature Pyramid Regression},
year={2024},
volume={},
number={},
pages={1-1},
keywords={Image stitching;Estimation;Thermal sensors;Deformation;Optical sensors;Image reconstruction;Accuracy;Multi-Spectral image stitching;image fusion;infrared and visible images;pyramid structure;deep learning},
doi={10.1109/TIP.2024.3430532}}
If you have any question or suggestion, please email to Zengxi Zhang or Zhiying Jiang.