A nonlinear fractional reaction-diffusion system applied to image denoising and decomposition

Abdelghafour Atlas; Mostafa Bendahmane; Fahd Karami; Driss Meskine; Omar Oubbih

doi:10.3934/dcdsb.2020321

Article Contents

2021, Volume 26, Issue 9: 4963-4998. Doi: 10.3934/dcdsb.2020321

This issue Previous Article On existence and uniqueness properties for solutions of stochastic fixed point equations Next Article A comparative study of atomistic-based stress evaluation

A nonlinear fractional reaction-diffusion system applied to image denoising and decomposition

1.
Ecole Nationale des Sciences Appliquées de Marrakech, Université Cadi Ayyad, B.P. 575 Avenue Abdelkrim Al Khattabi Marrakech, Morocco
2.
Institut de Mathématiques de Bordeaux and INRIA-Carmen Bordeaux Sud-Ouest, Université de Bordeaux, 33076 Bordeaux Cedex, France
3.
Ecole Supérieure de Technologie d'Essaouira, Université Cadi Ayyad, B.P. 383 Essaouira El Jadida, Essaouira, Morocco

^* Corresponding author: Fahd Karami
^* Corresponding author: Fahd Karami

Received: May 2020

Revised: July 2020

Early access: November 2020

Published: September 2021

Abstract / Introduction Full Text(HTML) Figure(15) / Table(3) Related Papers Cited by

Abstract

This paper is devoted to the mathematical and numerical study of a new proposed model based on a fractional diffusion equation coupled with a nonlinear regularization of the Total Variation operator. This model is primarily intended to introduce a weak norm in the fidelity term, where this norm is considered more appropriate for capturing very oscillatory characteristics interpreted as a texture. Furthermore, our proposed model profits from the benefits of a variable exponent used to distinguish the features of the image. By using Faedo-Galerkin method, we prove the well-posedness (existence and uniqueness) of the weak solution for the proposed model. Based on the alternating direction implicit method of Peaceman-Rachford and the approximations of the Gr$ \ddot{u} $nwald-Letnikov operators, we develop the numerical discretization of our fractional diffusion equation. Experimental results claim that our model provides high-quality results in cartoon-texture-edges decomposition and image denoising. In particular, our model can successfully reduce the staircase phenomenon during the image denoising. Furthermore, small details, texture and fine structures still maintained in the restored image. Finally, we compare our numerical results with the existing models in the literature.

Keywords:

Mathematics Subject Classification: 26A33, 35K57, 94A08, 46E3x.

Citation:

Full Text(HTML)

Figure 1. Comparison between the regularization terms of the function $ f(x) = 1/|x| $ associated to the TV operator, which is weak in the numerical simulations, (the lines associated to this function are illustrated by the red curves in both figures: case $ \epsilon = 0 $ and case $ \gamma = 0 $). The first figure illustrates the curves obtained by the function $ f(x) = 1/\sqrt{|x|^{2}+\epsilon} $ and by different values of $ \epsilon $. The second figure illustrates the curves obtained by $ f(x) = ( \mathop{{\rm{log}}}(1+|x|))^{\gamma}/|x| $ and by different values of $ \gamma $. It is clear that the previously proposed converges rapidly to the regularization term associated with the TV operator from $ \gamma = 10^{-1} $. On the other hand, the other function converges from $ \epsilon = 10^{-4} $

Download: Full-size image PowerPoint slide

Figure 2. Original images

Download: Full-size image PowerPoint slide

Figure 3. The efficiency test of our model to denoise corrupted images with $ \sigma = 10 $ (see Table 1)

Download: Full-size image PowerPoint slide

Figure 4. The efficiency test of our model to denoise corrupted images with $ \sigma = 20 $ (see Table 2)

Download: Full-size image PowerPoint slide

Figure 5. The efficiency test of our model to denoise corrupted images with $ \sigma = 30 $ (see Table 3)

Download: Full-size image PowerPoint slide

Figure 6. Image denoising performed on the medical image ('Lung' image). Noise removal results provided by our model and other denoising techniques

Download: Full-size image PowerPoint slide

Figure 7. Image denoising performed on the grayscale image ('Einstein' image). Noise removal results provided by our model and other denoising techniques

Download: Full-size image PowerPoint slide

Figure 8. Image denoising performed on the grayscale image ('Lena' image). Noise removal results provided by our model and other denoising techniques

Download: Full-size image PowerPoint slide

Figure 10. The comparison results of different values of the fraction $ s $ obtained by our proposed model. The first row contains smooth images $ u $; The second row contains a residual part associated with $ u-f $

Download: Full-size image PowerPoint slide

Figure 9. This figure shows the ability to reduce the staircase effect between restored images with our proposed model and the TV model using the $ L^{2}- $norm from [36]. Our model can successfully reduce the staircase phenomenon during the image denoising

Download: Full-size image PowerPoint slide

Figure 11. Comparisons results between the classical TV model ($ L^{2}- $norm) from [36] and the AAKM model ($ H^{-1}- $norm) from [3] with our proposed model ($ H^{-0.5}- $norm)

Download: Full-size image PowerPoint slide

Figure 12. Results on Scan's image of line profile number $ 50 $. Red: original image; Green: cartoon part obtained by the TV model using the $ L^{2}- $norm [36]; Violet: cartoon part obtained by the AAKM model using the $ H^{-1}- $norm [3]; Blue: cartoon part obtained by our proposed model using the $ H^{-0.5}- $norm

Download: Full-size image PowerPoint slide

Figure 13. Comparison results of edge detector, contour lines and texture applied on a cartoon image obtained by the classical TV model (using the $ L^{2}- $norm) from [36] and AAKM model (using the $ H^{-1}- $norm) from [3] with our proposed model (using the $ H^{-0.5}- $norm)

Download: Full-size image PowerPoint slide

Figure 14. Results on Pepper's image of line profile number $ 50 $. Red: original image; Green: cartoon part obtained by the TV model using the $ L^{2}- $norm from [36]; Violet: cartoon part obtained by the AAKM model using the $ H^{-1}- $norm from [3]; Blue: cartoon part obtained by our proposed model using the $ H^{-0.5}- $norm

Download: Full-size image PowerPoint slide

Figure 15. Comparison results of edge detector, contour lines and texture applied on a cartoon image obtained by the classical TV model (the $ L^{2}- $norm) from [36] and the AAKM model (the $ H^{-1}- $norm) from [3] with our proposed model (the $ H^{-0.5}- $norm)

Download: Full-size image PowerPoint slide

Table 1. The SNR and PSNR values of Noisy ($ \sigma=10 $) and Restored images of the first efficiency test

Noise level	$\underline{Fig.3(A)}$	$\underline{Fig.3(D)}$	$\underline{Fig.3(B)}$	$\underline{Fig.3(E)}$	$\underline{Fig.3(C)}$	$\underline{Fig.3(F)}$
$ \sigma=10 $	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$
$\mbox{SNR}$	12.54	20.23	13.57	14.84	12.66	15.11
$\mbox{PSNR}$	28.14	35.82	27.22	29.16	28.17	30.60
$\mbox{SSIM}$	0.782	0.971	0.865	0.933	0.590	0.949

| Show Table

DownLoad: CSV

Table 2. The $\mbox{SNR}$ and $\mbox{PSNR}$ values of Noisy ($ \sigma=20 $) and Restored images of the second efficiency test

$\mbox{Noise level}$	$\underline{Fig.4(A)}$	$\underline{Fig.4(D)}$	$\underline{Fig.4(B)}$	$\underline{Fig.4(E)}$	$\underline{Fig.4(C)}$	$\underline{Fig.4(F)}$
$ \sigma=20 $	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$
$\mbox{SNR}$	06.50	17.49	09.04	12.10	06.62	13.41
$\mbox{PSNR}$	22.10	33.09	22.02	25.38	22.13	28.90
$\mbox{SSIM}$	0.580	0.925	0.743	0.896	0.422	0.837

| Show Table

DownLoad: CSV

Table 3. The $\mbox{SNR}$ and $\mbox{PSNR}$ values of Noisy ($ \sigma=30 $) and Restored images of the third efficiency test

$\mbox{Noise level}$	$\underline{Fig.5(A)}$	$\underline{Fig.5(D)}$	$\underline{Fig.5(B)}$	$\underline{Fig.5(E)}$	$\underline{Fig.5(C)}$	$\underline{Fig.5(F)}$
$ \sigma=30$	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$	$\mbox{Noisy}$	$\mbox{Restored}$
$\mbox{SNR}$	03.01	16.12	05.51	10.07	03.06	12.44
$\mbox{PSNR}$	18.58	31.73	18.56	23.39	18.54	27.92
$\mbox{SSIM}$	0.501	0.884	0.650	0.885	0.325	0.780

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	R. Aboulaich, D. Meskine and A. Souissi, New diffusion models in image processing, Comput. Math. Appl., 56 (2008), 874-882. doi: 10.1016/j.camwa.2008.01.017.
[2]	R. A. Adams, Sobolev Spaces, Ac. Press, New york, 1975.
[3]	L. Afraites, A. Atlas, F. Karami and D. Meskine, Some class of parabolic systems applied to image processing, Discrete Contin. Dyn. Syst. Ser. B, 21 (2016), 1671-1687. doi: 10.3934/dcdsb.2016017.
[4]	L. Alvarez, P.-L. Lions and J.-M. Morel, Image selective smoothing and edge detection by nonlinear diffusion. Ⅱ, SIAM J. Numer. Anal., 29 (1992), 845-866. doi: 10.1137/0729052.
[5]	F. Andreu, C. Ballester, V. Caselles and J. M. Mazón, Minimizing total variation flow, Differential Integral Equations, 14 (2001), 321-360.
[6]	N. Aronszajn, Boundary values of functions with finite Dirichlet integral, Techn. Report of Univ. of Kansas, 14 (1955), 77-94.
[7]	G. Aubert and J.-F. Aujol, Modeling very oscillating signals. Application to image processing, Appl. Math. Optim., 51 (2005), 163-182. doi: 10.1007/s00245-004-0812-z.
[8]	J.-F. Aujol, G. Aubert, L. Blanc-Féraud and A. Chambolle, Image decomposition into a bounded variation component and an oscillating component, J. Math. Imaging Vision, 22 (2005), 71-88. doi: 10.1007/s10851-005-4783-8.
[9]	A. Buades, B. Coll and J. M. Morel, A non-local algorithm for image denoising, IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2 (2005), 60-65.
[10]	A. Chambolle, R. A. DeVore, N.-Y. Lee and B. J. Lucier, Nonlinear wavelet image processing: Variational problems, compression, and noise removal through wavelet shrinkage, IEEE Trans. Image Process., 7 (1998), 319–335. doi: 10.1109/83.661182.
[11]	E. A. Coddington and N. Levinson, Theory of ordinary differential equations, McGraw-Hill Book Company, Inc., New York-Toronto-London, (1955).
[12]	E. Di Nezza, G. Palatucci and E. Valdinoci, Hitchhiker's guide to the fractional Sobolev spaces, Bull. Sci. Math., 136 (2012), 521-573. doi: 10.1016/j.bulsci.2011.12.004.
[13]	K. Diethelm, The Analysis of Fractional Differential Equations, Lecture Notes in Mathematics, Springer-Verlag, Berlin, (2010). doi: 10.1007/978-3-642-14574-2.
[14]	S. Dipierro, X. Ros-Oton and E. Valdinoci, Nonlocal problems with Neumann boundary conditions, Rev. Mat. Iberoam., 33 (2017), 377-416. doi: 10.4171/RMI/942.
[15]	D. L. Donoho, De-noising by soft-thresholding, IEEE Transactions on Information Theory, 41 (1995), 613-627. doi: 10.1109/18.382009.
[16]	A. Elmahi and D. Meskine, Parabolic equations in Orlicz spaces, J. London Math. Soc. (2), 72 (2005), 410-428. doi: 10.1112/S0024610705006630.
[17]	A. Elmoataz, X. Desquesnes and O. Lézoray, Non-Local Morphological PDEs and $p$-Laplacian Equation on Graphs With Applications in Image Processing and Machine Learning, IEEE Journal of Selected Topics in Signal Processing, 6 (2012), 764-779. doi: 10.1109/JSTSP.2012.2216504.
[18]	E. Gagliardo, Ulteriori proprietà di alcune classi di funzioni in più variabili, Ricerche Mat., 8 (1959), 24-51.
[19]	J. B. Garnett, P. W. Jones, T. M. Le and L. A Vese, Modeling oscillatory components with the homogeneous spaces $B\dot MO^{-\alpha}$ and $\dot W{}^{-\alpha,p}$, Pure Appl. Math. Q., 7 (2011), 275-318. doi: 10.4310/PAMQ.2011.v7.n2.a2.
[20]	J. B. Garnett, T. M. Le, Y. Meyer and L. A Vese, Image decompositions using bounded variation and generalized homogeneous Besov spaces, Appl. Comput. Harmon. Anal., 23 (2007), 25-56. doi: 10.1016/j.acha.2007.01.005.
[21]	Y. Giga, M. Muszkieta and P. Rybka, A duality based approach to the minimizing total variation flow in the space $H^{-s}$, Jpn. J. Ind. Appl. Math., 36 (2019), 261-286. doi: 10.1007/s13160-018-00340-4.
[22]	G. Gilboa and S. Osher, Nonlocal linear image regularization and supervised segmentation, Multiscale Model. Simul., 6 (2007), 595-630. doi: 10.1137/060669358.
[23]	G. Gilboa and S. Osher, Nonlocal operators with applications to image processing, Multiscale Model. Simul., 7 (2008), 1005-1028. doi: 10.1137/070698592.
[24]	J.-P. Gossez, Nonlinear elliptic boundary value problems for equations with rapidly (or slowly) increasing coefficients, Trans. Amer. Math. Soc., 190 (1974), 163-205. doi: 10.1090/S0002-9947-1974-0342854-2.
[25]	Z. Guo, J. Yin and Q. Liu, On a reaction-diffusion system applied to image decomposition and restoration, Math. Comput. Modelling, 53 (2011), 1336-1350. doi: 10.1016/j.mcm.2010.12.031.
[26]	Y. Jin, J. Jost and G. Wang, A new nonlocal variational setting for image processing, Inverse Probl. Imaging, 9 (2015), 415-430. doi: 10.3934/ipi.2015.9.415.
[27]	Y. Kim and L. A. Vese, Image recovery using functions of bounded variation and Sobolev spaces of negative differentiability, Inverse Probl. Imaging, 3 (2009), 43-68. doi: 10.3934/ipi.2009.3.43.
[28]	S. Kindermann, S. Osher and P. W. Jones, Deblurring and denoising of images by nonlocal functionals, Multiscale Model. Simul., 4 (2005), 1091-1115. doi: 10.1137/050622249.
[29]	T. M. Le and L. A. Vese, Image decomposition using total variation and $ \rm{div} $($ \rm{BMO} $), Multiscale Model. Simul., 4 (2005), 390-423. doi: 10.1137/040610052.
[30]	L. H. Lieu and L. A. Vese, Image restoration and decomposition via bounded total variation and negative Hilbert-Sobolev spaces, Appl. Math. Optim., 58 (2008), 167-193. doi: 10.1007/s00245-008-9047-8.
[31]	X. Liu and L. Huang, A new nonlocal total variation regularization algorithm for image denoising, Math. Comput. Simulation, 97 (2014), 224-233. doi: 10.1016/j.matcom.2013.10.001.
[32]	Y. Meyer, Oscillating Patterns in Image Processing and Nonlinear Evolution Equations, American Mathematical Society, Providence, RI, (2001). doi: 10.1090/ulect/022.
[33]	K. S. Miller and B. Ross, An Introduction to the Fractional Calculus and Fractional Differential Equations, John Wiley & Sons, Inc., New York, (1993).
[34]	S. Osher, A. Solé and L. Vese, Image decomposition and restoration using total variation minimization and the $H^{-1}$ norm, Multiscale Model. Simul., 1 (2003), 349-370. doi: 10.1137/S1540345902416247.
[35]	P. Perona and J. Malik, Scale-space and edge detection using anisotropic diffusion, IEEE Transactions on Pattern Analysis and Machine Intelligence, 12 (1990), 629-639. doi: 10.1109/34.56205.
[36]	L. I. Rudin, S. Osher and E. Fatemi, Nonlinear total variation based noise removal algorithms, Phys. D., 60 (1992), 259-268. doi: 10.1016/0167-2789(92)90242-F.
[37]	L. N. Slobodeckiĭ, Generalized Sobolev spaces and their application to boundary problems for partial differential equations, Leningrad. Gos. Ped. Inst. Učen. Zap., 197 (1958), 54-112.
[38]	N. N. Tarkhanov, The Analysis of Solutions of Elliptic Equations, Kluwer Academic Publishers Group, Dordrecht, (1997). doi: 10.1007/978-94-015-8804-1.
[39]	L. A. Vese and S. J. Osher, Modeling textures with total variation minimization and oscillating patterns in image processing, J. Sci. Comput., 19 (2003), 553-572. doi: 10.1023/A:1025384832106.
[40]	Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, Image quality assessment: from error visibility to structural similarity, IEEE Transactions on Image Processing, 13 (2004), 600-612. doi: 10.1109/TIP.2003.819861.
[41]	L. P. Yaroslavsky, Digital Picture Processing, Springer-Verlag, Berlin, 1985. doi: 10.1007/978-3-642-81929-2.