Robust Watermarking through Gradient-Based Topological Encoding and Geometric Localization

This study introduces a robust image watermarking method that leverages the stability of topological structures.

This repository contains the implementation of the paper “Robust Watermarking through Gradient-Based Topological Encoding and Geometric Localization”, which has been accepted for publication in the Journal of Electronic Imaging. If you use this code in your research, please cite the paper.

Update Log

• [2025/12/18] README.
• [2025/08/01] code and README.

Requirements

• Python 3.6+
• OpenCV (cv2)
• NumPy
• SciPy
• scikit-image
• Matplotlib
• openpyxl

How to use

1:Prepare watermark and cover images Place watermark images such as HDU_W_20.png in the root directory. Include your test cover images (e.g., peppers.bmp, baboon.bmp) in the same folder.

2:Run the script

python code.py

3:Check outputs Results will be saved in folders like _Try/ Includes visualizations, extracted watermark images and logs PSNR, SSIM, NC, BER recorded in Try.xlsx

4:Attack Modes The following distortion attacks are applied during robustness evaluation: JPEG compression (Q=70, 50, 30) Image scaling (0.5×, 0.7×, 1.5×) Gaussian noise (σ=0.01, 0.03) Salt & pepper noise (amount=0.001, 0.01) Rotation Mean and median blurring Combined attacks (e.g., scale + noise + JPEG)

Citation

This repository provides the implementation of the paper “Robust Watermarking through Gradient-Based Topological Encoding and Geometric Localization”, which has been accepted for publication in the Journal of Electronic Imaging.  Please cite the paper if you use this code in your academic research.