GLF: A low-rank matrix based hyperspectral image denoiser

Hyperspectral Image Denoising Based on Global and Nonlocal Low-Rank Factorizations, ESI highly cited paper.

# HSI-denoiser-GLF

An HSI denoiser: Global Local Factorization (GLF)

The code and data herein distributed reproduce the results published in

the paper

Lina Zhuang and Jose M. Bioucas-Dias, "Hyperspectral image denoising based on global and non-local low-rank factorizations", IEEE Transactions on Geoscience and Remote Sensing, 2021. ESI highly cited paper, URL: https://ieeexplore.ieee.org/document/9318519

Lina Zhuang and Jose M. Bioucas-Dias, "Hyperspectral image denoising based on global and non-local low-rank factorizations", IEEE International Conference on Image Processing, Sep. 2017. URL: https://www.it.pt/Publications/DownloadPaperConference/30727

% Description:

main.m ---- main script reproducing the denoising results published in GLF paper

MPSNR.m ---- Performance criteria: mean peak signal-to-noise per band

MSSIM.m & ssim_index ---- Performance criteria: mean structural similarity per band

img_clean_dc.mat & img_clean_pavia.mat ---- Simulated clean datasets

GLF_denoiser.m ---- denoising algorithm GLF

NonlocalPatch_local_LR.m ---- Nonlocal patch-based denoising of subsapce coefficients

LowRank_tensor.m ---- Low-rank approximation based filtering for similar groups of patches

LowRankRecovery3.m ---- Low rank matrix recovery

hysime.m ---- Subsapce identification

% Notes:

1) Package instalation: unzip the files to a directory and run the

scripts of "main.m", which reproduces the denoising results

reported in the above paper.

2) The script GLF_denoiser.m use the functions (e.g., tensor()) from

Tensor Toolbox (version 2.6 (February 6, 2015)).

The MATLAB Tensor Toolbox Version 2.6 is available at the

web page: http://www.sandia.gov/~tgkolda/TensorToolbox/index-2.6.html

Download this toolbox and install it in the folder /tensor_toolbox

3) GLF_denoiser.m is the core funtion. It is a state-of-the-art denoiser

designed for hyperspectral images corrupted with additive Gaussian noise.

% ACKNOWLEDGMENTS

The authors acknowledge the following individuals and organisations:

- Prof. Paolo Gamba from Pavia university,

for making the Pavia University data set available to the community.

- Prof. David Landgrebe and Larry Biehl from Purdue University,

for making the Washington DC Mall data set available to the community.

- Authors of the MATLAB Tensor Toolbox (Brett W. Bader, Tamara G. Kolda,

Jimeng Sun, Evrim Acar, Daniel M. Dunlavy, Eric C. Chi, Jackson Mayo,

et al.) from Sandia National Laboratories, for making the Tensor

Toolbox available to the community.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Author: Lina Zhuang and Jose M. Bioucas Dias, Nov. 2017