/
README.txt
39 lines (33 loc) · 1.72 KB
/
README.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
This repository is made to publish code used in "Regularization by
Denoising: Clarifications and New Interpretations" by Reehorst, Schniter
https://arxiv.org/abs/1806.02296v2
This code can be used to recreate tables 1-3 and figures 1-9.
The figures and tables can be reproduced by their corresponding script:
tables 1-3 tables.m (change value of denoiser for the different
denoisers)
figure 1: cost_vs_fp_error.m
figure 2: RED_cost_visualization.m (change value of denoiser for the
different denoisers)
figure 3: ADMM_iteration_test.m
figure 4-6: algorithm_comparison.m (denoiser set to 'TNRD')
figure 7-9: algorithm_comparison.m (denoiser set to 'DWT')
This code uses several third-party code libraries. Directions on use and
links to the original code can be found below.
Using DnCNN:
DnCNN works much faster with matconvnet (http://www.vlfeat.org/matconvnet/)
Once matconvnet is installed add {matconvnet root folder}/matlab to the
matlab path and run vl_setupnn and make sure the useMatConvnet variable is
set to true. If matconvnet is compiled with GPU support, set useGPU = true,
else set to false. If you do not use matconvnet, set useMatConvnet = false
to use DnCNN
Using TNRD:
TNRD is only supported by windows and linux (not mac)
Using TDT (implemented using Rice-Wavelet-Toolbox):
Run rwt/bin/compile.m before using the rwt.
Links to original third-party code:
RED https://github.com/google/RED
BM3D http://www.cs.tut.fi/~foi/GCF-BM3D/index.html#ref_software
TNRD https://pan.baidu.com/s/1geNQP0J
DnCNN https://github.com/cszn/DnCNN
nlm https://www.mathworks.com/matlabcentral/fileexchange/52018-simple-non-local-means-nlm-filter
rwt https://github.com/ricedsp/rwt