convex_optimization

Blurring and deblurring algorithms.

Implementation of the following algorithms for image deblurring:

1. Primal Douglas-Rachford Splitting (Spingarn’s Method)
2. Primal-Dual Douglas Rachford Splitting
3. ADMM (Dual Douglas-Rachford)
4. Chambolle-Pock Method

Build instructions

Enter the convex_optimization directory and run: pip install -e .

Then you can import the functions in the project such as

from deblur_denoise.chambolle_pock.algorithm import chambolle_pock

Convex Optimization Algorithms

A collection of convex optimization algorithms for image deblurring and denoising, including:

1. Primal Douglas-Rachford Splitting (Spingarn’s Method)
2. Primal-Dual Douglas Rachford Splitting
3. ADMM (Dual Douglas-Rachford)
4. Chambolle-Pock Method

Installation

pip install -e .

Development Setup

Prerequisites

• Python 3.11.10
• pip
• Virtual environment (recommended)
• Required packages (from requirements.txt):
  • numpy>=1.21.0
  • scipy>=1.7.0
  • matplotlib>=3.4.0
  • torch (for GPU acceleration - TBA)

Project Structure

convex_optimization/
├── deblur_denoise/
│   ├── admm/              # ADMM algorithm implementation
│   │   ├── algorithm.py
│   │   └── __init__.py
│   ├── primal_dr/         # Primal Douglas-Rachford implementation
│   │   ├── algorithm.py
│   │   └── __init__.py
│   ├── primal_dual_dr/    # Primal-Dual Douglas-Rachford implementation
│   │   ├── algorithm.py
│   │   └── __init__.py
│   ├── chambolle_pock/    # Chambolle-Pock implementation
│   │   ├── algorithm.py
│   │   └── __init__.py
│   ├── core/              # Core functionality
│   │   ├── blur.py
│   │   ├── convolution.py
│   │   ├── loss.py
│   │   ├── noise.py
│   │   ├── proximal_operators.py
│   │   └── __init__.py
│   ├── op_math/           # Mathematical operators
│   │   ├── python_code/
│   │   ├── matlab_code/
│   │   └── __init__.py
│   ├── utils/             # Utility functions
│   │   ├── conv_utils.py
│   │   ├── logging_utils.py
│   │   ├── sample_images/
│   │   └── __init__.py
│   ├── blur_deblur.py     # Main deblurring functionality
│   └── __init__.py
├── .vscode/               # IDE configuration
│   ├── launch.json
│   └── tasks.json
├── build_test_script.py   # Development and testing script
├── setup.py               # Package configuration
├── requirements.txt       # Project dependencies
├── LICENSE                # MIT License
├── Project.md             # Project documentation
└── README.md              # This file

Development Workflow

1. Install in Development Mode

   pip install -e .

2. Running Tests You can run tests in several ways:

   a. Using the build task (recommended):

   • Press Cmd+Shift+B (Mac) or Ctrl+Shift+B (Windows/Linux)
   • This will automatically:
     • Uninstall the current package
     • Reinstall it in development mode
     • Run the test suite
     • The environment variable CO_IMAGE_PATH can be set to specify the image path for testing

   b. Using the command line:

   python build_test_script.py

   c. Using the debugger:

   • Open the Run and Debug view (Cmd+Shift+D or Ctrl+Shift+D)
   • Select "Python: Reinstall and Test"
   • Press F5 to start debugging

Test Suites

The development script tests all algorithms for a sample image whose path can be read from the environment variable CO_IMAGE_PATH.

Development Tools

• Build System: Uses setup.py for package management
• Testing: Custom test suite with metrics and visualization
• Debugging: VS Code/Cursor configuration for easy debugging

Best Practices

1. Always run the build task after making changes
2. Check test output for any errors
3. Use the debugger for complex issues
4. Keep dependencies up to date in requirements.txt
5. Use proper tensor operations (clone/detach) when working with PyTorch tensors
6. Document any changes to algorithm parameters

Pulling Changes

To update a specific file from master:

git checkout master -- path/to/file

License

MIT License

Authors

• Alexander Kakabadze
• Edmand Yu
• Lilian Yuan
• Rahul Padmanabhan