MCAN

This repository is the official pytorch implementation of "Mosaic Convolution-Attention Network for Demosaicing Multispectral Filter Array Images" (TCI 2021) (https://ieeexplore.ieee.org/abstract/document/9507356)

Requirements

• PyTorch (1.0)

Usage

Training

You can download the training dataset from https://pan.baidu.com/s/1S_kaYiDWJS7iz-oF_oHx1A, the password is 5cul. usage: main_MCAN.py

Evaluation using CAVE

usage: eval.py

Demo for synthetic spectral mosaic image

usage: demo.py

Demo for real spectral mosaic image

usage: demo_realimage.py

Citation

If you find the code and datasets useful in your research, please cite:

@article{feng2021mosaic, title={Mosaic Convolution-Attention Network for Demosaicing Multispectral Filter Array Images}, author={Feng, Kai and Zhao, Yongqiang and Chan, Jonathan Cheung-Wai and Kong, Seong G and Zhang, Xun and Wang, Binglu}, journal={IEEE Transactions on Computational Imaging}, volume={7}, pages={864--878}, year={2021}, publisher={IEEE} }