- Officical code of paper "Deep Neural Network based Sparse Measurement Matirx for Image Compressed Sensing" ICIP2018
- Download the paper: https://arxiv.org/pdf/1806.07026v1.pdf
- Windows10
- Matlab R2015b
- MatconvNet 1.0-beta23 (https://www.vlfeat.org/matconvnet/)
- CUDA 8.0
- Generating the imdb training data for DSMM by running
GenerateData_DSMM.m
- Train the DSMM, run the code
Demo_Train_DSMM.m
Matrix_Sparse.m
is used for sparse constraint and Matirx_orth.m
is used for norm constraint.
- Obtain the learned measurement matrix from the trained model by running
get_sampling_mat.m
- Using the learned measurement matrix to replace the traditional gaussian matrix for different iteration based CS methods.
- For training data, you can choose any dataset by yourself.
- If you like this repo, Star or Fork to support my work. Thank you.
- If you have any problem, please email wxcui@hit.edu.cn
- If you find the code is useful in your research, please cite:
@article{Cui2018An,
title={Deep Neural Network based Sparse Measurement Matirx for Image Compressed Sensing},
journal={IEEE International Conference on Image Processing (ICIP)},
author={Cui, Wenxue, Jiang, Feng, Gao, Xinwei, Tao, Wen and Zhao, Debin},
year={2018},
}
This code is built based on the repo https://github.com/cszn/DnCNN