Requirements

• Download 3DVideoDenoising.zip at https://hulaa.ddns.net/nextcloud/index.php/s/B9HPRPrGRzXCCxn

• Extract this zip into the root directory of this repo

• MATLAB R2019b

• MatConvNet version 1.0-beta24

• CUDA release 10.1.

• cuDNN v7.6.5

• Optional: For plotting maximally perceptually-distinct colors in graph validation: https://de.mathworks.com/matlabcentral/fileexchange/29702-generate-maximally-perceptually-distinct-colors

Config

In the main folder the config.m enables configuration of the denoising process:

1. Setup path to repository
2. Setup path to input .vol files
3. Setup path to input .mat files
4. Define region of interest (slices)

If you denoise a .vol file for the first time, this software will convert the .vol to a .mat file and perform denoising on the .mat file. If you already converted the .vol file and perform a second denoising run, you can simply set useMATfile to true and specify the path of the .mat file.

Denoising a volume

1. Run the script run.m
2. The denoising results are stored in the folder results/ along with the denoised data and runtime needed.

Denoising methods

1. Gauss
2. Median
3. Mean Box
4. Laplacian of Gaussian
5. Block Matching 3D (BM3D)
6. Guided Image Filtering
7. Total Variation Denoising
8. Non local means Filter
9. Wiener
10. Wavelet denoising
11. neighShrink SURE
12. Gauss 3D
13. Mean Box 3D
14. Median 3D
15. DnCNN
16. DnCNN MATLAB
17. FFD net
18. IRCNN
19. WavResNet
20. WavResNet RNN
21. AAPM-Challenge
22. MWCNN

About

Hauptseminar Advanced Image Processing - WS19/20 - Universität Stuttgart

Marcel Hasenbalg & Seda Ulusal