SRAD - Speckle Reducing Anisotropic Diffusion

A Python implementation of the Speckle Reducing Anisotropic Diffusion (SRAD) algorithm for denoising images affected by speckle noise, commonly found in ultrasound, SAR (Synthetic Aperture Radar), and other coherent imaging systems.

---

📖 About

This package is a Python port of the original MATLAB implementation by Xingorno. The SRAD algorithm is based on anisotropic diffusion, which smooths homogeneous regions while preserving edges — making it ideal for speckle reduction without blurring important structures.

How It Works

SRAD extends the classic Perona-Malik anisotropic diffusion by incorporating a speckle-sensitive diffusion coefficient. The algorithm iteratively applies a diffusion process where:

• Homogeneous regions (high speckle) → Strong smoothing
• Edge regions (low speckle, high gradient) → Preserved

The mathematical model follows:

$$\frac{\partial I}{\partial t} = \text{div}[c(q) \nabla I]$$

Where $c(q)$ is the diffusion coefficient computed from the instantaneous coefficient of variation $q$.

---

🚀 Installation

From PyPI

pip install srad

From Source

git clone https://github.com/Keno-00/Speckle-Reducing-Anisotropic-Diffusion-SRAD.git
cd Speckle-Reducing-Anisotropic-Diffusion-SRAD
pip install .

---

💻 Usage

As a Library

import cv2
import srad

# Load your grayscale image
img = cv2.imread('path/to/image.png', cv2.IMREAD_GRAYSCALE)

# SRAD Parameters
iteration_max = 200    # Number of diffusion iterations
time_step = 0.05       # Time step per iteration (stability: keep ≤ 0.25)
decay_factor = 1       # Controls how fast edges are "locked in"

# Apply SRAD denoising
denoised = srad.SRAD(img, iteration_max, time_step, decay_factor)

# Save result
cv2.imwrite('denoised.png', denoised)

Command Line

Run the built-in example on noisyImage.png:

python -m srad

---

⚙️ Parameters

Parameter	Type	Description	Recommended
iterationMaxStep	int	Maximum number of diffusion iterations	100-300
timeSize	float	Time step size (affects stability and speed)	0.05-0.15
decayFactor	float	Exponential decay for the diffusion coefficient	1.0

Tips:

• Higher iterationMaxStep → More smoothing, longer runtime
• Larger timeSize → Faster convergence, but may become unstable if > 0.25
• Higher decayFactor → Edges lock in faster

---

📁 Repository Structure

.
├── srad/                    # Python package
│   ├── __init__.py          # SRAD algorithm implementation
│   └── __main__.py          # CLI entry point
├── SpeckleReducingAD.m      # Original MATLAB implementation (general)
├── SpeckleReducingAD_New.m  # Original MATLAB implementation (optimized)
├── testSRAD.m               # MATLAB test script
├── noisyImage.png           # Example input image
├── denoised.png             # Example output image
├── setup.py                 # Package configuration
├── pyproject.toml           # Build system configuration
├── LICENSE                  # MIT License
└── README.md                # This file

---

📚 References

This implementation is based on the following papers:

1. Yu, Y., & Acton, S. T. (2002). Speckle reducing anisotropic diffusion. IEEE Transactions on Image Processing, 11(11), 1260-1270.
   📄 IEEE Xplore

2. Perona, P., & Malik, J. (1990). Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(7), 629-639.
   📄 IEEE Xplore

---

🙏 Acknowledgments

• Original MATLAB Implementation: Xingorno
• Python Port: Keno S. Jose (this contribution)
• Initial Python Contribution: namioj

This project is a fork of the original repository, ported to Python for easier integration into image processing pipelines.

---

📄 License

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

The original MATLAB implementation by Xingorno is also under the MIT License.