Flex-DLD: Deep Low-rank Decomposition Model with Flexible Priors for Hyperspectral Image Denoising and Restoration
- Denoising Data Preparation
- Datasets for hyperspectral images denoising experiments include: KAIST, Washington DC Mall, CAVE, Indian Pines
You can download in: Google drive or Baidu drive (Code: emg2)
Put the downloaded data into the [Data] folder
Alternatively, you can generate noisy hyperspectral images according to the code
- Denosing your hyperspectral images
Put your data into the [Data] folder
The mat file should include [noisy_img] and [img] variables
- Denoising Experiments
cd Denoise
python main_*.py
(We have uploaded KAIST Scene01 data. You can run main_1_KAIST.py directly.)
- Check Our Denoising Results
The denoised hyperspectral images of KAIST, WDC, CAVE, and IP datasets are available here: Baidu drive (Code: 4g02)
- Restoration Data Preparation
Datasets for hyperspectral images restoration experiments include: KAIST
You can download in: Google drive or Baidu drive (Code: u4er)
Put the downloaded data into the [Data] folder
- Restoration Experiments
cd Denoise/CASSI_Restoration/
python main.py
(We have uploaded KAIST Scene01 data. You can run main.py directly.)
- Check Our Reconstructed Results
The reconstructed hyperspectral images of KAIST datasets are available here: Google drive or Baidu drive (Code: 87hh)
- Code Description
main_*.py : code for denoising
optimization.py : code of ADMM iteration
func.py : code includes some useful functions
test_metric : code for evaluation
model/model_loader.py : code for loading deep low-rank networks
model/LRNet.py : code for designing network architecture
model/common.py : code includes some network blocks
model/basicblock.py : code includes some network blocks
model/utils.py : code includes functions for evaluation