International Journal of Discoveries and

Innovations in Applied Sciences

| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8

A Noise Density-Based Fuzzy Approach for Detecting and Removing

Random Impulse Noise in Color Images

Adel Jalal Yousif

University of Diyala, Diyala, Iraq

Abstract:
This paper introduces a new approach aimed at restoring images corrupted by random valued
impulse noise. The adopted methodology leverages fuzzy logic and encompasses three primary
stages: estimation of noise density, detection of fuzzy noise, and reduction of fuzzy noise. Within
the fuzzy noise detection phase, a fuzzy set labeled as "Noise-Free" is formulated through the
utilization of the rank-ordered mean of absolute differences and the estimated noise density. This
set serves to discern whether a given pixel should be classified as noisy or noise-free. Utilizing the
fuzzy logic in the proposed method collaborates to determine the ultimate fuzzy weight assigned to
each pixel, thereby facilitating the restoration of corrupted image pixels. Empirical results based on
peak signal-to-noise ratio, mean square error, and visual assessment demonstrate the effectiveness
of the proposed technique in suppressing noise, preserving fine details, and surpassing the
performance of several established filtering methods.

1. Introduction
In numerous digital image processing applications, the captured images often suffer from noise
corruption, stemming from issues during image transmission or acquisition. Noise degradation
diminishes image quality and introduces adverse effects into subsequent image processing stages,
including segmentation, parameter estimation, and enhancement. Consequently, the elimination of
noise from images emerges as a fundamental and critical task within the domain of image
processing. Impulse noise, characterized by brief 'on/off' noise pulses of relatively short duration,
holds a significant and pervasive presence among digital image noise sources. Its impact is felt
during image acquisition due to factors like noisy sensors (arising from switching or sensor
temperature variations), as well as during transmission due to channel imperfections (interference,
atmospheric disturbances), hardware-related defects, or synchronization errors during analog-to-
digital conversion within image processing workflows. This study specifically focuses on the
Random Valued Impulse Noise (RVIN) model of impulse noise [1-5].
Among the diverse techniques for mitigating impulse noise, the median filter [6,7] has gained
popularity due to its effective noise suppression capabilities coupled with high computational
efficiency. However, the median filter often suffers from drawbacks such as destroying image
details and introducing blurring, as each pixel is replaced with the median value of its surrounding
neighborhood. To address this issue, several forms of adaptive and center-weighted median filter
[8-15] was introduced, which assigns greater emphasis to the central pixel within the observed
window. Although this modification preserves more image details than the median filter, it
uniformly treats all pixels across the image, without distinguishing between noisy and noise-free
pixels. This paper introduces a new approach for removing RVIN from images. The proposed
method employs fuzzy techniques alongside estimated noise density. The estimation of noise
density serves dual purposes: first, to determine the appropriate window size for the fuzzy noise
detection step, and second, to shape the membership function used in the fuzzy noise detection
process. The fuzzy logic, which is adopted by Zadeh in 1965, is a mathematical tool for dealing
Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

1
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
with uncertainty. It presents a crucial model of computing with words and furnishes an approach
for handling imprecision and information granularity within the realm of soft computing [16-20].
2. RVIN Model
Impulse noise inherently maintains its independence and lacks correlation with individual image
pixels, resulting in a subset of pixels bearing noise while the remainder remains untarnished. When
dealing with random valued impulse noise, colloquially termed uniform impulse noise, the affected
pixels can adopt any value within the gray level spectrum, denoted as [ ] in an 8-bit
image representation. Consequently, these noisy pixels span the full gray level range from 0 to 255.
This scenario entails a stochastic dispersal of noise throughout the entire image, leading to an equal
probability of any given gray level value manifesting as noise. The Random Valued Impulse Noise
(RVIN) model can be briefly described as follows [21, 22]:
𝑦
… (1)
𝑦 {
𝑦

Where, ( ,𝑦) is the noisy image pixel, ( ,𝑦) is the noisy impulsive pixel at position ( ,𝑦), and
( ,𝑦) is the uncorrupted (original) image pixel.
3. Proposed Method
The proposed method consists of four main concepts: estimation of noise density, noise detection,
noise removing, and iterative processing which are described below in details.
3.1. Estimation of Noise Density
For estimating RVIN density in color image, each channel (color component) is processed
separately by dividing it into 16 non-overlapping blocks (Bi) with i ∈ {1,2,…,16} and then the
noise density for each block in each channel is calculated as described below for the first block
only but it is performed in an analogous way for the others:
1- Calculating the standard deviation for the entire block based on the following
equation:

√ ̀ ̀
∑ (̀ 𝑦 ) ... (2)
∈

where, : is the mean value of first block

̀ 𝑦 : is the pixel data inside .
2- Scanning the whole image block using a 3x3 sliding window from pixel to pixel and using
the following three steps in each scan:
i) |̀ 𝑦 ̀ 𝑦 | ∈ …(3)
Where, represents the set of absolute differences between the central pixel and its four nearest
neighbors within the processed window and
denotes the set of coordinates for the four neighbors of 𝑦 given by:
{ } ... (4)

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

2
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
Then the elements in can be arranged in ascending order by value:

ii) Defining which represents the mean value of the second and third ordered elements in
by the following equation:

... (5)
iii) Using the standard deviation of the processed image block and the value of to
identify the current pixel 𝑦 as follows:

𝑦 { ... (6)

Where, 𝑦 denotes the decision rule for estimating RVIN density.

3- The noise density for the first block is given as:

∑ ∈ 𝑦
... (7)
̀ ̀

Where ̀ ̀ denotes the size of the block , and denotes the control factor which is
estimated by experiments (see table 1).
4- The calculation of the noise density for the other image blocks is done in the same way as
for . Then, define the minimum value of as:

( ) …(8)
, with ∈ { }
5- If is less than one, then the image is classified as clean. Otherwise, the image is
classified as corrupted with RVIN. In this case, the overall noise density of the image equals to
the mean value of the noise densities across all blocks and expressed is as follows:

{ ... (9)
( )
Where, is the RVIN density for the entire channel.
3.2. Nosie Detection
As stated in section 2, in the context of RVIN, a noisy pixel holds the potential to exhibit any value
within the gray level range [0-255] for an 8-bit image representation. Additionally, it might slightly
deviate in intensity from the original value. Consequently, mitigating such noise presents a greater
challenge compared to dealing with salt and pepper noise, necessitating a robust detection strategy.
The proposed method employs the Rank-Ordered Mean of the Absolute Differences (ROMAD) in
conjunction with estimated noise density, integrating the principles of fuzzy logic. The process of
calculating the rank-ordered mean difference involves the central pixel's comparison with its
neighboring pixels within a sliding window of dimensions (2𝐾+1)×(2𝐾+1), encompassing each
pixel in the image. The value of 𝐾 is dynamically determined based on the prevailing noise density:
Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

3
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
𝐾 is set to 1 when ( RV<35%) and otherwise, it takes on the value of 2. To compute (ROMAD)
for a given pixel ( ,𝑦), the following three steps are executed:
1- Compute the absolute differences, denoted as between the current pixel 𝑦 and
neighboring pixels within the observed window using the following formula :
| 𝑦 𝑦 | ∈{ 𝐾 𝐾} ... …(10)
Where, 𝑦 denotes the pixels values within the processed window.
2- Assume be the smallest ranked value in when the elements of are arranged in
ascending order, such that

Where, is the number of the central pixel neighbors in the processed window 𝐾 𝐾
. So, 𝐾
3- The (ROMAD) for the current pixel is defined as:
∑
... (11)
Where, 𝐾 ... (12)
Upon completing the calculation of ROMAD for every pixel within the image, the ROMAD value
serves as a valuable metric for discerning between pixels that are affected by noise and those that
remain noise-free. Specifically, in scenarios where the noise ratio is low and the ROMAD value for
a given pixel I(x,y) is small, it signifies that the pixel is noise-free and exists within a homogenous
neighborhood. Conversely, a larger ROMAD value suggests that the pixel may be noisy or situated
on an edge. Consequently, a fuzzy set labeled "Small ROMAD" is introduced to capture the
linguistic notion of "small," and its membership function, denoted as and illustrated in figure
1, is employed to identify noise-free pixels. However, a challenge arises when dealing with high
noise ratios, leading to consistently elevated ROMAD values. To address this issue, the
membership function for the "Small ROMAD" fuzzy set can be dynamically shaped (i.e., its
parameters determined) for each neighborhood. This adaptability is governed by two key criteria:
firstly, the homogeneity level of the neighborhood surrounding the tested pixel I(x,y).
Secondly, the overall noise density of the entire image.

Fig.1: Membership function of fuzzy set “Small ROMAD”

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

4
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
The following equation describing the membership function :

( ) ... (13)

{
Where:
... (14)
... (15)
denotes the noise factor and represents the nth smallest value in .
Further, the following steps describing how , and can be obtained:
Firstly, can be computed by utilizing the ROMAD values in the observed window as:
1- Assume be the smallest ranked value in when the elements of are
arranged in ascending order, such that:

Where, 𝐾
2- obtain the value of as:
∑
... (16)
Secondly, the value of q is obtained as:
... (17)
Thirdly, the value of can be calculated by utilizing the straight line equation as illustrated in
figure 2 and using the following expression:
... (18)

Fig. 2: The function of

Where the parameters ( 1, 2, and 3) are determined through experimental investigations.
Empirical findings have indicated that the optimal values for these parameters are set to (2.9, 0, 61)
respectively when the estimated noise density remains at or below 35%. Conversely, when the
estimated noise density surpasses 35%, the parameters ( 1, 2, and 3) are optimized at values of
(1.9, 20, 100) respectively.
Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

5
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
Lastly, for each tested pixel 𝑦 , a fuzzy set “Noise-Free” is generated using the following
fuzzy rule:
Fuzzy Rule 1: Defined when a central pixel 𝑦 is a noise free pixel:
IF is small
THEN the central pixel is noise- free
This rule can be applied by employing the equality operation between two fuzzy sets. As a result,
the membership function of the fuzzy set “Noise-Free” is calculated as:
𝑦 ( ) ... (19)
3.3 Noise Removing
The noise removing step is demonstrated for the red component only (i.e., ) but it is implemented
in an analogous way for the other components. Hence, assume that 𝑦 is a noisy pixel
(i.e. 𝑦 )) and 𝑦 is the corresponding pixel of the filtered red
component, and then one of the following cases will be applied for restoring the noisy pixel as
follows:
Case1: IF ( 𝑦 𝑦 )
𝑦 𝑦 𝑦 … (20)
Case2: IF ( 𝑦 𝑦 )
𝑦 𝑦 𝑦 … (21)
Case3: IF ( 𝑦 𝑦 )
𝑦 ( ( ( 𝑦 𝑦 𝑦 𝑦 ) ) ) … (22)
Case4: IF ( 𝑦 𝑦 )
𝑦
∑ ∑
( 𝑦 ) ∑ ∑
( 𝑦 𝑦 ) … (23)
Where, 𝑦 represents the pixels values in the considered window of size 𝐾
𝐾 . The size of observed window is selected adaptively according to the number of the noise
free pixel in that window starting with . If the observed window is fully noisy, then the size
of window will be increased until the condition ( ) is met.

∑ ∑ 𝑦 ( 𝑦 ) … (24)

3.4 Iterative Processing

During the noise detection stage, the membership functions , as illustrated in figure 1,
dynamically adjust their shape based on the estimated noise density. Consequently, the proposed
algorithms are executed in an iterative manner with two iterations. This iterative approach aims to
enhance noise removal while preserving the details of image. In the second iteration, the altered
image obtained from the initial iteration is utilized
Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

6
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
4. Experimental Results
Identifying a suitable parameter is a very significant task to obtain successful filtering results.
Hence, different experiments have been implemented to select the best values for the control factor
Eq. (7) which has a significant effect on the result of the noise density estimation. Table 1
contain different noise levels with parameters varying over a range of possible values for the
“Peppers” image. It is clear from table 1 that the best value of the parameter CF equal to 1.35,
where the difference between the estimated noise density and the real noise density is less than 5%.
Hence, the suggested value is =1.35.
Table 1: Determination of parameter CF on “Peppers” image
Color Parameter Real Noise Density (RVIN)
components values 10% 20% 30% 40% 50% 60%
0.95 7.8% 15.0% 22.0% 29.2% 35.4% 40.1%
1.15 9.4% 18.1% 26.7% 35.3% 42.9% 48.6%
Red
1.35 11.0% 21.3% 31.3% 41.5% 50.4% 57.0%
component
1.55 12.7% 24.5% 35.9% 47.6% 57.8% 65.5%
1.75 14.3% 27.6% 40.6% 53.8% 65.3% 73.9%
0.95 5.9% 12.5% 19.4% 25.9% 33.0% 38.8%
1.15 7.1% 15.2% 23.6% 31.4% 39.9% 46.9%
Green
1.35 8.4% 17.8% 27.7% 36.9% 46.9% 55.1%
component
1.55 9.7% 20.5% 31.8% 42.3% 53.9% 63.3%
1.75 10.9% 23.1% 35.9% 47.8% 60.8% 71.5%
0.95 7.2% 15.1% 22.4% 29.3% 35.8% 40.7%
1.15 8.8% 18.3% 27.2% 35.5% 43.4% 49.3%
Blue
1.35 10.3% 21.5% 31.9% 41.7% 50.9% 57.9%
component
1.55 11.9% 24.7% 36.7% 47.8% 58.5% 66.5%
1.75 13.4% 27.8% 41.4% 54.0% 66.0% 75.0%

The performance evaluation of the suggested method involves a comparison with several existing
noise reduction methods: DWM [23], HFC [24], CAFSM [25], and FMVMF [26]. Furthermore, the
same proposed method is applied to grayscale images without incorporating the color differences
step, termed as proposed (gray) in this section. Table 2 presents the quantitative outcomes of
objective quality measurements, specifically Mean Squared Error (MSE) and Peak Signal-to-Noise
Ratio (PSNR) [27-30], for the “Peppers” image (refer to figure 4). This image is subjected to
corruption by 10%, 30%, and 50% RVIN. It is obvious from table 1 that the proposed color image
algorithm has a superior performance as compared with the other noise reduction methods and with
the proposed gray scale image algorithm as well. The impact of noise density on the proposed
algorithm's performance, as well as the performances of related methodologies in terms of PSNR, is
visualized in figure 3. While, figure 4 illustrates the noise filtering and detail preservation of the
proposed method as compared with other methods. It is obvious from figure 4 that the proposed
color image algorithm obtains the best result in low and high noise densities due to the following
main reasons:
1. Adaptive Window Size: The utilization of an adaptive window size during both the noise
detection and noise filtering stages is informed by the number of noise-free pixels present
within the observed window.

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

7
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
2. Color Component Correlation: Exploiting the inter-correlation between color components
during the noise filtering stage. So, for restoring a certain noisy pixel 𝑦 at position 𝑦 ,
the corresponding pixels in the other components i.e., ( 𝑦 and 𝑦 ) are used instead
of the neighbors pixels for 𝑦 .
3. Robust Fuzzy Noise Detection: The integration of a robust and powerful fuzzy noise detection
scheme further contributes to the exceptional performance of the suggested method.
Table 2: Numerical results of the suggested method with related works using “Peppers”
image.
Noise density
10% 30% 50%
Method
MSE MSE MSE
PSNR PSNR PSNR
(102) (102) (102)
Noisy 18.30 1.4932 13.43 4.5994 11.15 7.7794
FMVMF 30.85 0.0822 21.53 0.7111 16.00 2.5806
HFC 29.89 0.1036 17.75 1.6909 12.94 5.1493
CAFSM 28.32 0.1528 26.04 0.2556 22.67 0.5519
DWM 31.40 0.0726 26.65 0.2164 22.44 0.5753
Proposed
34.81 0.0332 29.48 0.1146 25.62 0.2778
(gray)
Proposed
37.24 0.0191 30.82 0.0840 26.24 0.2397
(color)

Fig. 3: Comparison chart in term of PSNR of the proposed color image algorithm with
related works using the “Peppers” image corrupted with RVIN (10% - 60%).

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

8
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8

Fig. 4: Results of RVIN removing of “Peppers” image, (a) original image, (b) Noisy image
corrupted with 30 % RVIN (PSNR: 13.43), (c) Processed image with HFC (PSNR: 17.75), (d)
Processed image with FMVMF (PSNR: 21.53), (e) Processed image with CAFSM (PSNR:
26.04), (f) Processed image with DWM (PSNR: 26.65) (g) Processed image with the proposed
gray image algorithm (PSNR: 29.48), (h) Processed image using the proposed color image
algorithm (PSNR: 30.82).
5. Conclusion
This study suggests a new method for RVIN detection and reduction method for color image. The
proposed method utilizes fuzzy logic, noise density estimation, and correlation between RGB
Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

9
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
components. The method capitalizes on the calculated noise density to dynamically determine an
optimal window size for noise detection and shape the membership function employed in the noise
detection process. Through comprehensive experimentation, the proposed method showcases
exceptional performance surpassing numerous established impulse noise filtering techniques in
both objective and subjective evaluations. Notably, the method excels in effectively removing noise
while simultaneously preserving intricate image textures and details.
References
1. J. Chen, G. Zhang, S. Xu and H. Yu, "A Blind CNN Denoising Model for Random-Valued
Impulse Noise," in IEEE Access, vol. 7, pp. 124647-124661, 2019.
2. C. Lin, Y. Li, S. Feng and M. Huang, "A Two-Stage Algorithm for the Detection and Removal
of Random-Valued Impulse Noise Based on Local Similarity," in IEEE Access, vol. 8, pp.
222001-222012, 2020.
3. S. K. Maji and A. Mahajan, "A Joint Denoising Technique for Mixed Gaussian–Impulse Noise
Removal in HSI," in IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1-5, 2023.
4. P. H. Sangave and G. P. Jain, "Impulse noise detection and removal by modified boundary
discriminative noise detection technique," 2017 International Conference on Intelligent
Sustainable Systems (ICISS), Palladam, India, 2017, pp. 715-719.
5. T. M. Y. Shiju and A. V. N. Krishna, "A Two-Pass Hybrid Mean and Median Framework for
Eliminating Impulse Noise From a Grayscale Image," 2021 2nd International Conference on
Advances in Computing, Communication, Embedded and Secure Systems (ACCESS),
Ernakulam, India, 2021, pp. 206-210.
6. W. K. Pratt, “Median filtering,” Image Proc. Institute, University of Southern California, Los
Angeles, Tech. Rep., September 1975.
7. S. Deivalakshmi, S. Sarath and P. Palanisamy, "Detection and removal of Salt and Pepper noise
in images by improved median filter," 2011 IEEE Recent Advances in Intelligent
Computational Systems, Trivandrum, India, 2011, pp. 363-368, doi:
10.1109/RAICS.2011.6069335.
8. S.J. Ko, Y.H. Lee, “Center weighted median filters and their applications to image
enhancement”, IEEE Trans. Circuits and Systems, Vol. (38), pp. (984–993), 1991
9. A. Sindhu and V. Radha, "A Novel Histogram Equalization Based Adaptive Center Weighted
Median Filter for De-noising Positron Emission Tomography (PET) Scan Images," 2018 3rd
International Conference on Communication and Electronics Systems (ICCES), Coimbatore,
India, 2018, pp. 909-914, doi: 10.1109/CESYS.2018.8724108.
10. T. Chen and Hong Ren Wu, "Adaptive impulse detection using center-weighted median filters,"
in IEEE Signal Processing Letters, vol. 8, no. 1, pp. 1-3, Jan. 2001, doi: 10.1109/97.889633.
11. T. Rabie, "Hybrid mean adaptive center weighted median filter for color sensor denoising,"
Proceedings of 2004 International Symposium on Intelligent Signal Processing and
Communication Systems, 2004. ISPACS 2004., Seoul, Korea (South), 2004, pp. 321-326, doi:
10.1109/ISPACS.2004.1439068.
12. F. J. Gallegos-Funes, B. E. Carvajal-Gamez, J. L. Lopez-Bonilla and V. I. Ponomaryov,
"Steganographic method based on wavelets and center weighted median filter," 2010
INTERNATIONAL KHARKOV SYMPOSIUM ON PHYSICS AND ENGINEERING OF

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

10
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
MICROWAVES, MILLIMETER AND SUBMILLIMETER WAVES, Kharkiv, Ukraine, 2010,
pp. 1-3, doi: 10.1109/MSMW.2010.5546011.
13. S. Salekin, S. Agaian, I. Jahan and S. Z. Sajal, "Image De-Noising Through Symmetric, Bell-
Shaped, and Centered Weighted Median Filters Based Subband Decomposition," 2018 IEEE
International Conference on Electro/Information Technology (EIT), Rochester, MI, USA, 2018,
pp. 0493-0497, doi: 10.1109/EIT.2018.8500238.
14. V. Oliinyk, O. Ieremeiev and I. Djurović, "Center Weighted Median Filter Application to Time
Delay Estimation in Non-Gaussian Noise Environment," 2019 IEEE 2nd Ukraine Conference
on Electrical and Computer Engineering (UKRCON), Lviv, Ukraine, 2019, pp. 985-989, doi:
10.1109/UKRCON.2019.8879775.
15. A. Taguchi and N. Izawa, "Fuzzy center weighted median filters," 1996 8th European Signal
Processing Conference (EUSIPCO 1996), Trieste, Italy, 1996, pp. 1-4.
16. M. M. Hamid, F. Fathi Hammad and N. Hmad, "Removing the Impulse Noise from Grayscaled
and Colored Digital Images Using Fuzzy Image Filtering," 2021 IEEE 1st International
Maghreb Meeting of the Conference on Sciences and Techniques of Automatic Control and
Computer Engineering MI-STA, Tripoli, Libya, 2021, pp. 711-716.
17. V. Narayan, P. K. Mall, S. Awasthi, S. Srivastava and A. Gupta, "FuzzyNet: Medical Image
Classification based on GLCM Texture Feature," 2023 International Conference on Artificial
Intelligence and Smart Communication (AISC), Greater Noida, India, 2023, pp. 769-773.
18. P. V. S. Reddy, "Generalized Fuzzy Logic with twofold fuzzy set: Learning through Neural Net
and Application to Business Intelligence," 2021 International Conference on Fuzzy Theory and
Its Applications (iFUZZY), Taitung, Taiwan, 2021, pp. 1-5.
19. N. Kamide, "Sequential Fuzzy Description Logic: Reasoning for Fuzzy Knowledge Bases with
Sequential Information," 2020 IEEE 50th International Symposium on Multiple-Valued Logic
(ISMVL), Miyazaki, Japan, 2020, pp. 218-223.
20. H. M. Rehan Afzal, J. Yu and Y. Kang, "Impulse noise removal using fuzzy logics," 2017 32nd
Youth Academic Annual Conference of Chinese Association of Automation (YAC), Hefei,
China, 2017, pp. 413-418.
21. Dr. Muna M. Jawad, Dr. Ekbal H. Ali, Adel J. Yousif, “A Fuzzy Random Impulse Noise
Detection and Reduction Method Based on Noise Density Estimation”, International Journal of
Scientific & Engineering Research, Volume 5, Issue 3, March-2014.
22. C. Anjanappa and H. S. Sheshadri, "Development of mathematical morphology filter for
medical image impulse noise removal," 2015 International Conference on Emerging Research
in Electronics, Computer Science and Technology (ICERECT), Mandya, India, 2015, pp. 311-
318, doi: 10.1109/ERECT.2015.7499033.
23. Y. Dong and S. Xu, “A New Directional Weighted Median Filter for Removal of Random-
Valued Impulse Noise”, IEEE Signal Processing Letters, vol. (14), No. (3), pp. (193-196),
March, 2007.
24. S. Schulte, V. De Witte, M. Nachtegael, D. Van Der Weken, E. E. Kerre, “Histogram-Based
Fuzzy Colour Filter for Image Restoration”, Image and Vision Computing, Vol. (25), pp.
(1377-1390), 2007.

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

11
 International Journal of Discoveries and
Innovations in Applied Sciences
| e-ISSN: 2792-3983 | www.openaccessjournals.eu | Volume: 3 Issue: 8
25. J.Harikiran, B.Saichandana, B.Divakar “Impulse Noise Removal in Digital Images”
International Journal of Computer Applications, Vol. (11), No. (8), pp. (39-42), November,
2010.
26. B. Smolka, M. Szczepanski, K.N. Plataniotis, and A.N. Venetsanopoulos, “Fast Modified
Vector Median Filter”, Computer Analysis of Images and Patterns, LNCS 2124, pp. (570-580),
2001.
27. Adel Jalal Yousif, “Image Steganography Based on Wavelet Transform and Color Space
Approach”, Diyala Journal of Engineering Sciences Vol (13) No 3, 2020: 23-34.
28. Fadhil Kadhim Zaidan Ghazwan Jabbar Ahmed, Adel Jalal Yousif, “A Digital Image
Watermarking Scheme based on Discrete Cosine Transform”, Journal of Engineering and
Applied Sciences, vol. 4, issue 16, pp. 5762-5768, 2019.
29. Adel Jalal Yousif, “Image Steganography Based on Wavelet Transform and Color Space
Approach”, Diyala Journal of Engineering Sciences Vol (13) No 3, 2020: 23-34.
30. J. Moreno, B. Jaime and C. Fernandez, "NRPSNR: No-Reference Peak Signal-to-Noise Ratio
for JPEG2000," 2013 Data Compression Conference, Snowbird, UT, USA, 2013, pp. 511-511,
doi: 10.1109/DCC.2013.117.

Published under an exclusive license by open access journals under Volume: 3 Issue: 8 in Aug-2023
Copyright (c) 2023 Author (s). This is an open-access article distributed under the terms of Creative Commons
Attribution License (CC BY).To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

12