Richardson-Lucy deconvolution algorithm with PyTorch

This repository provides a GPU implementation of the Richardson-Lucy deconvolution algorithm leveraging the computational capabilities of PyTorch. The Richardson-Lucy algorithm is widely used for image deblurring and restoration, particularly in the context of microscopy and astronomical imaging.

Examples

Non-blind deconvolution with a greyscale image

Blind deconvolution with a greyscale image

Non-blind deconvolution with a color image

Installation

Prerequisites

• Anaconda or Miniconda installed on your system

Environment Setup

You can create the conda environment using the environment.yaml file.

1. Clone this repository to your local machine.
2. Navigate to the repository's root directory in your terminal.
3. Create a Conda environment using the provided environment.yaml file:

conda env create -f environment.yaml

Usage

The core functionality is provided by the richardson_lucy function. To use it, import the function from richardson_lucy.py and pass your image along with the point spread function (PSF) and other parameters as needed.

from richardson_lucy import richardson_lucy

deconvolved_image = richardson_lucy(y, x, k, steps=20, tv=False)

You can also find an usage example in the greyscale.ipynb and color.ipynb jupyter notebooks. To test the algorithm, several types of blur are implemented such as Gaussian blur and motion blur (thanks to the code from this repository). TV regularization is also implemented using a torch implementation of Chambolle-Pock denoising. Color images are only partially supported, for instance TV regularization does not work yet.

License

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