Python wrapper for BM3D for stationary correlated noise (including white noise) for color, grayscale and multichannel images and deblurring.
BM3D is an algorithm for attenuation of additive spatially correlated stationary (aka colored) Gaussian noise. This package provides a wrapper for the BM3D binaries for use for grayscale, color and other multichannel images for denoising and deblurring.
This implementation is based on
- Y. Mäkinen, L. Azzari, A. Foi, 2020, "Collaborative Filtering of Correlated Noise: Exact Transform-Domain Variance for Improved Shrinkage and Patch Matching", in IEEE Transactions on Image Processing, vol. 29, pp. 8339-8354.
- K. Dabov, A. Foi, V. Katkovnik, K. Egiazarian, 2007, "Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering", in IEEE Transactions on Image Processing, vol. 16, pp. 2080-2095.
This package provides a BM3D interface for the "bm4d" denoising package. Please see the bm4d package for supported platforms.
The package is available for non-commercial use only. For details, see LICENSE.
For examples, see the examples folder of the full source (bm3d-***.tar.gz) from https://pypi.org/project/bm3d/#files , which also includes the example noise cases demonstrated in the paper.
Authors:
Ymir Mäkinen ymir.makinen@tuni.fi
Lucio Azzari
Alessandro Foi