Skip to content

GeorgeKAHChen/HOSVD_denoise_patch

Repository files navigation

Image denoising based on global image similar patches searching and HOSVD to patches tensor

Cite this project/paper

Guo, J., Chen, H., Shen, Z. et al. Image denoising based on global image similar patches searching and HOSVD to patches tensor. EURASIP J. Adv. Signal Process. 2022, 19 (2022). https://doi.org/10.1186/s13634-021-00798-4

Install

git clone https://www.github.com/KazukiAmakawa/HOSVD_denoise_patch

Usage

Easy Mode

Detemine the parameters in main.m and run the project

Parameter Table

parameter Intro
para_sigma Sigma of noise
para_betta Relaxation parameter (Learning Rate)
para_gamma Scaling factor controlling
para_patch_size Size of every patch
para_patch_stack Length of the tensor in SVD processing
para_iteration Iteration times
test_switch Print the image on screen or not
patch_method Patch Analysis method
  • We had found best para_betta and para_gamma in sigma = 10, 30 and 50.

  • We just trained the GMM pre-trained model in 7, 8, 9 and 10 patch size

  • patch_method list

Code Intro
1 Original method NNM patch search
21/22 Pre-trained Gaussian Mixture Model method only
31/32/33/34/35 Pre-trained GMM and K-means method
21/31 BFS after classification search
22/32 Virtual reference patch method
33 All patch in class combined tensor
34 Reference patch in classes to build tensor
35 Double port search on image

Full Mode

You can using HOSVD_Denoising.m directly for your denoising.

Or, add new patch methods in Block_matching.m to test your new patch method with HOSVD

Source Code: https://www.github.com/KazukiAmakawa/HOSVD
For more help, submit issue in this project or connect this E-mail: GeorgeKahChen@gmail.com

LICENSE

Copyright (c) by KazukiAmakawa(Huayan Chen), all right reserved.
GNU GENERAL PUBLIC LICENSE Version 3

If you want to using this project as close source project, please connect us.

Reference

[1] GARW: Group Attribute Random Walk: https://www.github.com/KazukiAmakawa/GARW-Class

(This is a project working on non-linear classify method in deep learning with pytorch structure, still developing)

[2] F. Chen, L. Zhang and H. Yu, "External Patch Prior Guided Internal Clustering for Image Denoising," 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, 2015, pp. 603-611. doi: 10.1109/ICCV.2015.76

(We used GMM and K-means solution from code of this paper)

[3] R. Movchan and Z. Shen, "Adaptive thresholding hosvd algorithm with iterative regularization for image denoising," 2017 IEEE International Conference on Image Processing (ICIP), Beijing, 2017, pp. 2991-2995. doi: 10.1109/ICIP.2017.8296831

Contact

If you have any problem about the code, please contact: (Mr.) Huayan Chen: GeorgeKahChen@gmail.com

About

HOSVD denoising algorithm with different patch method

Resources

Stars

Watchers

Forks

Packages

No packages published