SAR Image Despeckling Using Continuous Attention Module

Model Architecture

Installation

Clone this repo.

git clone https://github.com/JK-the-Ko/SAR-CAM.git
cd SAR-CAM/

Please install dependencies by

python3 -m pip install -r requirements.txt

Usage

Train

python3 train.py --project PROJECT_NAME --noisy-train-dir NOISY_IMAGE_TRAIN_DIR --clean-train-dir CLEAN_IMAGE_TRAIN_DIR --noisy-valid-dir NOISY_IMAGE_VALID_DIR --clean-valid-dir CLEAN_IMAGE_VALID_DIR 

Inference

python3 test.py --weights-dir SAVE_WEIGHT_DIR --clean-image-dir CLEAN_IMAGE_TEST_DIR --noisy-image-dir NOISY_IMAGE_TEST_DIR --save-dir DENOISED_IMAGE_SAVE_DIR

Despeckling Performance

Results for the freeway image with 1-look speckle noise. (a) Reference. (b) Noisy image. (c) PPB. (d) SAR-BM3D. (e) FANS. (f) SAR-DRN. (g) HDRANet. (h) U-Net. (i) STD-CNN. (j) MONet. (k) MRDDANet. (l) Proposed Method., respectively.

Results for the parking lot image with 1-look speckle noise. (a) Reference. (b) Noisy image. (c) PPB. (d) SAR-BM3D. (e) FANS. (f) SAR-DRN. (g) HDRANet. (h) U-Net. (i) STD-CNN. (j) MONet. (k) MRDDANet. (l) Proposed Method., respectively.

Citation

If you use SAR-CAM in your work, please consider citing us as

@ARTICLE{9633208,
  author={Ko, Jaekyun and Lee, Sanghwan},
  journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing}, 
  title={SAR Image Despeckling Using Continuous Attention Module}, 
  year={2022},
  volume={15},
  number={},
  pages={3-19},
  doi={10.1109/JSTARS.2021.3132027}}