IJCSNS International Journal of Computer Science and Network Security, VOL.2 No.

ISSN 1254-6354, Feb 2023

Image De-noising using Discrete Wavelet

transform
B. Deena Divya Nayomi, B.Soujanya

Sep 22 ,2022

Summary
simplicity, it is frequently used in the literature. The main
The image de-noising naturally corrupted by noise is a classical
problem in the field of signal or image processing. Additive
idea is to subtract the threshold value T from all wavelet
random noise can easily be removed using simple threshold coefficients larger than T, arising from the standard
methods. De-noising of natural images corrupted by Gaussian discrete wavelet transform and to set all other coefficients
noise using wavelet techniques are very effective because of its to zero.
ability to capture the energy of a signal in few energy transform
values. The wavelet de-noising scheme thresholds the wavelet The problem of Image de-noising can be
coefficients arising from the standard discrete wavelet summarized as follows. Let A(i,j) be the noise-free image
transform. In this paper, it is proposed to investigate the and B(i,j)the image corrupted with independent Gaussian
suitability of different wavelet bases and the size of different noise Z(i,j),
neighborhood on the performance of image de-noising
algorithms in terms of PSNR.
B(i,j)= A(i,j)+σ Z(i,j) ……(1)
Key words:
Image, De-noising, Wavelet, Transform
where Z(i,j) has normal distribution N(0,1). The
problem is to estimate the desired signal as accurately as
1. Introduction possible according to some criteria. In the wavelet
domain, if an orthogonal wavelet transform is used, the
This paper investigates the suitability of different problem can be formulated as
wavelet bases and the size of different neighborhood on Y(i,j)= W(i,j)+N(i,j) ……(2)
the performance of image de-noising algorithms in terms where Y(i,j) is noisy wavelet coefficient; W(i,j) is
of PSNR. Over the past decade, wavelet transforms have true coefficient and N(i,j) noise, which is independent
received a lot of attention from researchers in many Gaussian. . In this paper, it is proposed to investigate the
different areas. Both discrete and continuous wavelet suitability of different wavelet bases and the size of
transforms have shown great promise in such diverse different neighborhood on the
fields as image compression, image de-noising, signal performance of image de-noising algorithms in terms of
processing, computer graphics, and pattern recognition to PSNR.
name only a few. In de-noising, single orthogonal
wavelets with a single-mother wavelet function have
played an important role. De-noising of natural images 2. Discrete Wavelet transform
corrupted by Gaussian noise using wavelet techniques is
very effective because of its ability to capture the energy The Discrete Wavelet Transform (DWT) of image
of a signal in few energy transform values. Crudely, it signals produces a non-redundant image representation,
states that the wavelet transform yields a large number of which provides better spatial and spectral localization of
small coefficients and a small number of large image formation, compared with other multi scale
coefficients. Simple de-noising algorithms that use the representations such as Gaussian and Laplacian pyramid.
wavelet transform consist of three steps. Recently, Discrete Wavelet Transform has attracted more
 Calculate the wavelet transform of the and more interest in image de-noising. The DWT can be
noisy signal. interpreted as signal decomposition in a set of
independent, spatially oriented frequency channels. The
 Modify the noisy wavelet coefficients
signal S is passed through two complementary filters and
according to some rule.
emerges as two signals, approximation and Details. This
 Compute the inverse transform using the
is called decomposition or analysis. The components can
modified coefficients.
be assembled back into the original signal without loss of
One of the most well-known rules for the second
information. This process is called reconstruction or
step is soft thresholding. Due to its
synthesis. The mathematical manipulation, which implies
effectiveness and
analysis and synthesis, is called discrete wavelet transform
 IJCSNS International Journal of Computer Science and Network Security, VOL.2 No.ISSN 1254-6354, Feb 2023

and inverse discrete wavelet transform. An image can be

decomposed into a sequence of different spatial 3. Wavelet Based ImageDe-noising
resolution images using DWT. In case of a 2D image, an
All digital images contain some degree of noise. Image
N level decomposition can be performed resulting in
de- noising algorithm attempts to remove this noise from
3N+1 different frequency bands namely, LL, LH, HL and
the image. Ideally, the resulting de-noised image will not
HH as shown in figure 1. These are also known by other
contain any noise or added artifacts. De-noising of
names, the sub-bands may be respectively called a 1 or the
natural images corrupted by Gaussian noise using
first average image, h1 called horizontal fluctuation, v1
wavelet techniques is very effective because of its ability
called vertical fluctuation and d1 called the first diagonal
to capture the energy of a signal in few energy transform
fluctuation. The sub-image a 1 is formed by computing the
values. The methodology of the discrete wavelet
trends along rows of the image followed by computing
transform based image de-noising has the following three
trends along its columns. In the same manner,
steps as shown in figure 2. 1. Transform the noisy image
fluctuations are also created by computing trends along
into orthogonal domain by discrete 2D wavelet
rows followed by trends along columns. The next level of
transform. 2. Apply hard or soft thresholding the noisy
wavelet transform is applied to the low frequency sub
detail coefficients of the wavelet transform 3. Perform
band image LL only. The Gaussian noise will nearly be
inverse discrete wavelet transform to obtain the de-noised
averaged out in low frequency wavelet coefficients.
image. Here, the threshold plays an important role in the
Therefore, only the wavelet coefficients in the high
de- noising process. Finding an optimum threshold is a
frequency levels need to be thresholded.
tedious process. A small threshold value will retain the
noisy coefficients whereas a large threshold value leads
to the loss of coefficients that carry image signal details.
Normally, hard thresholding and soft thresholding
techniques are used for such de-noising process. Hard
thresholding is a keep or kill rule whereas soft
thresholding shrinks the coefficients above the threshold
in absolute value. It is a shrink or kill rule.

Figure 1: 2D-DWT with 3-Level decomposition

Figure2: Diagram of wavelet based image De-noising

The following are the methods of threshold selection for

rest wavelet coefficients are very small. This algorithm
image de-noising based on wavelet transform
offers the advantages of smoothness and adaptation.
However, it exhibits visual artifacts.
Method 1: Visushrink
Threshold T can be calculated using the Method 2: Neighshrink
formulae, T= σ√2logn2 …….(3)
Let d(i,j) denote the wavelet coefficients of
This method performs well under a number of
interest and B(i,j) is a neighborhood window around
applications because wavelet transform has the
d(i,j). Also let S2=∑d2(i,j) over the window B(i,j).
compaction property of having only a small number of
Then the
large coefficients. All the
 IJCSNS International Journal of Computer Science and Network Security, VOL.2 No.ISSN 1254-6354, Feb 2023

wavelet coefficient to be thresholded is shrinked according evaluate the different de-noising scheme like Wiener
to the formulae, filter, Visushrink, Neighshrink and Modified
d(i,j)= d(i,j)* B(i,j) ….(4) Neighshrink.
where the shrinkage factor can be defined as B(i,j) = ( 1-
T2/ S2(i,j))+, and the sign + at the end of the formulae 5. Experiments
means to keep the positive value while set it to zero when
it is negative. Quantitatively assessing the performance in practical
application is complicated issue because the ideal image
is normally unknown at the receiver end. So this paper
Method 3: Modineighshrink uses the following method for
experiments. One original image is applied with Gaussian
During experimentation, it was seen that when the noise with different variance. The methods proposed for
noise content was high, the reconstructed image using implementing image de-noising using wavelet transform
Neighshrink contained mat like aberrations. These take the following form in general. The image is
aberrations could be removed by wiener filtering the transformed into the orthogonal domain by taking the
reconstructed image at the last stage of IDWT. The cost wavelet transform. The detail wavelet coefficients are
of additional filtering was slight reduction in sharpness of modified according to the shrinkage algorithm. Finally,
the reconstructed image. However, there was a slight inverse wavelet is taken to reconstruct the de-noised
improvement in the PSNR of the reconstructed image image. In this paper, different wavelet bases are used in
using wiener filtering. The de-noised image using all methods. For taking the wavelet transform of the
Neighshrink sometimes unacceptably blurred and lost image, readily available MATLAB routines are taken. In
some details. The reason could be the suppression of too each sub-band, individual pixels of the image are shrinked
many detail wavelet coefficients. This problem will be based on the threshold selection. A de-noised wavelet
avoided by reducing the value of threshold itself. So, the transform is created by shrinking pixels. The inverse
shrinkage factor is given by wavelet transform is the de-noised image.

6. Results and Discussions

B(i,j) = ( 1- (3/4)*T2/ S2(i,j))+ …….(5)
For the above mentioned three methods, image de-
noising is performed using wavelets from the second
4. Evaluation Criteria
level to fourth level decomposition and the results are
The above said methods are evaluated using the quality shown in figure (3) and table if formulated for second
measure Peak Signal to Noise ratio which is calculated level decomposition for different noise variance as
using the formulae, follows. It was found that three level decomposition and
fourth level decomposition gave optimum results.
PSNR= 10log 10 (255) 2/MSE (db) … (6) However, third and fourth level decomposition resulted
where MSE is the mean squared error between the in more blurring. The experiments were done using a
original image and the reconstructed de-noised image. It window size of 3X3, 5X5 and 7X7. The neighborhood
is used to window of 3X3 and 5X5 are good choices.
 IJCSNS International Journal of Computer Science and Network Security, VOL.2 No.ISSN 1254-6354, Feb 2023

Figure3: Results of various Image De-noising Methods

7. Conclusion
In this paper, the image de-noising using discrete wavelet transform is analyzed. The experiments were conducted to study
the suitability of different wavelet bases and also different window sizes. Among all discrete wavelet bases, coiflet
performs well in image de-noising. Experimental results also show that modified Neighshrink gives better result than
Neighshrink, Weiner filter and Visushrink.

References
[1] D. L. Donoho, “Denoising by soft-thresholding,” IEEE Transactions on Information Theory, vol. 41, pp.613-627, 1995.
[2] Bui and G. Y. Chen, “Translation invariant de-noising using multi- wavelets,” IEEE Transactions on Signal Processing, vol.46, no.12, pp.3414-3420,
1998.
[3] L. Sendur and I. W. Selesnick, “Bi-variate Shrinkage with Local Variance Estimation,” IEEE Signal Processing Letters, Vol. 9, No. 12, pp. 438-
441,2002.
[4] G. Y. Chen and T. D. Bui, “Multi-wavelet De-noising using Neighboring Coefficients,” IEEE Signal Processing Letters, vol.10, no.7, pp.211-214,
2003.
[5] M. A. Hannan, M. M. Hoque, A. Hussain, Y. Yusof, and P. J. Ker, “State-of-the-art and energy management system of lithium- ion batteries in
electric vehicle applications: issues and recommendationsreferences,” IEEE Access, vol. 6, pp. 19362–19378, 2018, doi:
10.1109/ACCESS.2018.2817655.
[6] M. Pahlevani and P. Jain, “A fast DC-bus voltage controller for bidirectional single-phase AC/DC converters,” IEEE Transactions on Power
Electronics, vol. 30, no. 8, pp. 4536–4547, 2015, doi: 10.1109/TPEL.2014.2356413.
[7] M. Azeroual, T. Lamhamdi, H. El Moussaoui, and H. El Markhi, “Simulation tools for a smart grid and energy management for microgrid with
wind power using multi-agent system,” Wind Engineering, vol. 44, no. 6, pp. 661–672, Dec. 2020, doi: 10.1177/0309524X19862755.
[8] S. A. Gorji, H. G. Sahebi, M. Ektesabi, and A. B. Rad, “Topologies and control schemes of bidirectional DC–DC power converters: an
overview,” IEEE Access, vol. 7, pp. 117997–118019, 2019, doi: 10.1109/ACCESS.2019.2937239.
[9] M. J. Er and Y. L. Sun, “Hybrid fuzzy proportional-integral plus conventional derivative control of linear and nonlinear systems,”
IEEE Transactions on Industrial Electronics, vol. 48, no. 6, pp. 1109–1117, 2001, doi: 10.1109/41.969389.
[10] B. Fernández R. and J. K. Hedrick, “Control of multivariable non-linear systems by the sliding mode method,” International Journal of
Control, vol. 46, no. 3, pp. 1019–1040, Sep. 1987, doi: 0.1080/00207178708547410.
[11] Y. Boujoudar, M. Azeroual, H. El Moussaoui, and T. Lamhamdi, “Intelligent controller based energy management for stand‐alone power
system using artificial neural network,” International Transactions on Electrical Energy Systems, vol. 30, no. 11, Nov. 2020, doi: 10.1002/2050-
7038.12579.
[12] M. A. G. de Brito, L. Galotto, L. P. Sampaio, G. de A. e Melo, and C. A. Canesin, “Evaluation of the main MPPT techniques for photovoltaic
applications,” IEEE Transactions on Industrial Electronics, vol. 60, no. 3, pp. 1156–1167, Mar. 2013, doi: 10.1109/TIE.2012.2198036.
[13] R. Sakipour and H. Abdi, “Voltage stability improvement of wind farms by self-correcting static volt-ampere reactive compensator and energy
storage,” International Journal of Electrical Power & Energy Systems, vol. 140, Sep. 2022, doi: 10.1016/j.ijepes.2022.108082.
[14] G. Chaspierre, P. Panciatici, and T. Van Cutsem, “Control of a battery energy storage system to compensate for ADN equivalents
inaccuracies,” Electric Power Systems Research, vol. 213, no. 6, Dec. 2022, doi: 10.1016/j.epsr.2022.108455.
[15] C. Parthasarathy, H. Khajeh, H. Firoozi, H. Laaksonen, and H. Hafezi, “Distributed generation, storage and active network management,” in
Deregulated Electricity Structures and Smart Grids, 1st ed., Boca Raton: CRC Press, 2022, pp. 83–111, doi: 10.1201/9781003278030-5.
[16] S. Kichou, T. Markvart, P. Wolf, S. Silvestre, and A. Chouder, “A simple and effective methodology for sizing electrical energy storage (EES)
systems based on energy balance,” Journal of Energy Storage, vol. 49, no. 11, May 2022, doi: 10.1016/j.est.2022.104085.
[17] N. S. Luu, K.-Y. Park, and M. C. Hersam, “Characterizing and mitigating chemomechanical degradation in high-energy lithium- ion battery
cathode materials,” Accounts of Materials Research, vol. 3, no. 5, pp. 511–524, 2022, doi: 10.1021/accountsmr.1c00282.
[18] Y. Boujoudar, H. Elmoussaoui, and T. Lamhamdi, “Lithium-ion batteries modeling and state of charge estimation using artificial neural
network,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 5, pp. 3415–3422, Oct. 2019, doi:
10.11591/ijece.v9i5.pp3415-3422.
[19] M. Shafiee-Rad, M. S. Sadabadi, Q. Shafiee, and M. Reza Jahed-Motlagh, “Modeling and robust structural control design for hybrid AC/DC
microgrids with general topology,” International Journal of Electrical Power & Energy Systems, vol. 139, no. 4, Jul. 2022, doi:
10.1016/j.ijepes.2022.108012