FDWM - Frequency Domain Watermarking

A Python library for embedding and extracting watermarks in images using frequency domain techniques. Supports both image and text watermarks with a command-line interface.

Features

• Frequency Domain Watermarking: Uses FFT (Fast Fourier Transform) to embed watermarks in high-frequency regions
• Grid-based Embedding: Embeds watermark in multiple grid cells for robust extraction
• Image Watermarks: Embed and extract image-based watermarks
• Text Watermarks: Embed and extract text watermarks with OCR support
• Command Line Interface: Easy-to-use CLI for batch processing
• Multiple Languages: Support for various text languages
• Robust Extraction: High correlation coefficients for reliable watermark detection

Installation

From PyPI (Recommended)

pip install fdwm

From Source

git clone https://github.com/Liam0205/fdwm.git
cd fdwm
pip install -e .

Quick Start

Python API

import cv2
from fdwm import embed, extract, extract_text

# Load host image and watermark
host_img = cv2.imread('host.jpg')
watermark_img = cv2.imread('watermark.png')

# Embed watermark using grid strategy (3x3 grid by default)
out_path, metrics = embed(host_img, watermark_img, strength=0.1)
print("Watermarked image saved to:", out_path)
print("Metrics:", metrics)

# Extract watermark
extracted_watermark = extract(out_path, watermark_img.shape[:2])

# For text watermarks
text = "Hello World"
out_path, metrics = embed(host_img, None, 'watermarked.png', watermark_text=text, strength=0.1)
extracted_text = extract_text(out_path)

Note:

• embed returns a tuple (output_path, metrics).
• metrics is a dict with keys: 'mean_pixel_diff', 'max_pixel_diff', 'p90_pixel_diff', 'psnr'.

Command Line Interface

# Embed image watermark (3x3 grid by default)
fdwm embed host.jpg --watermark-img watermark.png

# Embed with custom grid size
fdwm embed host.jpg --watermark-img watermark.png --grid-m 4 --grid-n 4

# Embed text watermark
fdwm embed host.jpg --watermark-text "Hello World"

# Extract image watermark
fdwm extract watermarked.jpg

# Extract text watermark
fdwm extract watermarked.jpg --text

Embedding Method

FDWM uses a grid-based embedding strategy:

1. Grid Division: The image is divided into m × n grid cells (default: 3×3)
2. Cell Processing: Each grid cell is processed independently in the frequency domain
3. Corner Embedding: Within each cell, the watermark is embedded in the top-left corner
4. Fusion: During extraction, all grid cells are processed and averaged for robust recovery

This approach provides excellent robustness against cropping, rotation, and other geometric attacks.

CLI Usage

Embed Command

fdwm embed <host_image> [options]

Options:
  --watermark-img PATH    Watermark image path
  --watermark-text TEXT   Watermark text
  --strength FLOAT        Embedding strength (default: 30000.0)
  --scale FLOAT           Watermark scale relative to host (default: 0.25)
  --font PATH             Font file path for text watermark
  --font-size INT         Font size for text watermark
  --grid-m INT            Number of vertical grid divisions (default: 3)
  --grid-n INT            Number of horizontal grid divisions (default: 3)
  --debug                 Print detailed metrics for each processed image

Extract Command

fdwm extract <watermarked_image> [options]

Options:
  --strength FLOAT        Embedding strength used during embedding (default: 30000.0)
  --scale FLOAT           Watermark scale relative to host (default: 0.25)
  --output PATH           Directory to save extracted watermark images/text
  --text                  Perform OCR and output text instead of image
  --save-text             Save recognized text to .txt files
  --grid-m INT            Number of vertical grid divisions used during embedding (default: 3)
  --grid-n INT            Number of horizontal grid divisions used during embedding (default: 3)

Examples

Basic Usage

# Embed watermark with default 3x3 grid
fdwm embed image.jpg --watermark-text "Secret"

# Embed with 4x4 grid for higher robustness
fdwm embed image.jpg --watermark-text "Secret" --grid-m 4 --grid-n 4

# Extract text
fdwm extract watermarked.jpg --text

Batch Processing

# Embed watermark in all images in a directory
fdwm embed images/ --watermark-img watermark.png

# Extract watermark from all images
fdwm extract watermarked/ --output extracted/

Text Watermarking

# Embed Chinese text
fdwm embed image.jpg --watermark-text "Hello World"

# Extract text
fdwm extract watermarked.jpg --text

Requirements

• Python 3.10+
• numpy
• opencv-python
• Pillow
• pytesseract

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments

• Based on frequency domain watermarking techniques