471
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
AN ENHANCEMENT OF EDGE PRESERVATION
IN OAMNHA DENOISING USING TEXTURE
BOUNDARIES
Indupriya Kumarasamy
Research Scholar, Department of Computer Science, Sri Ramakrishna College of
Arts and Science, Coimbatore, Tamilnadu (India).
E-mail: indupriya1406@gmail.com
ORCID: https://orcid.org/0000-0003-0446-9490
Anna Saro Vijendran
Dean, School of Computing, Sri Ramakrishna College of
Arts and Science, Coimbatore, Tamilnadu (India).
E-mail: saroviji@redimail.com
ORCID: https://orcid.org/0000-0003-2200-1226
Recepción: 28/11/2019 Aceptación: 31/03/2021 Publicación: 30/11/2021
Citación sugerida:
Kumarasamy, I., y Vijendran, A. S. (2021). An enhancement of edge preservation in OAMNHA
denoising using texture boundaries. 3C Tecnología. Glosas de innovación aplicadas a la pyme, Edición Especial,
(noviembre, 2021), 471-489. https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
472 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
ABSTRACT
In image processing, the most challenging task is removing the noise from the images since
the segmentation of image patches from noisy images has high complexity. To tackle this
challenge, an Optimum Adaptive parameterized Mask Non-Harmonic Analysis (OAMNHA)
based image denoising technique was proposed in Curvelet Transform (CT) domain. In this
technique, the image was converted into frequency domain representation to decompose
it into dierent subband. Then, edge-preserving segmentation using canny edge detection
was applied for each subband to extract the edges and identify the edge regions from a noisy
image. However, this process has high time consumption due to its complex computation.
Hence in this article, a Neuro-Fuzzy (NF) methodology is proposed as edge-preserving
segmentation method. Initially, noisy images represented in the frequency domain are given
to the NF edge detector to segment the edge regions and homogeneous textures from a
noisy image. Moreover, OAMNHA technique is applied for each region excluding edge
regions to restore the noiseless image accurately. This segmentation of noisy images covers
Neuro-fuzziness in the choice of the region boundary. Based on this segmentation, the
boundary distortion is eciently minimized since it denes texture boundaries with less
time consumption and computational complexity. Also, the accuracy of edge-preserving
segmentation is increased signicantly. Finally, the experimental results prove that the
proposed NF-OAMNHA-CT based image denoising technique has better performance
than the OAMNHA-CT technique.
KEYWORDS
Image denoising, Edge-preserving segmentation, Canny edge detection, OAMNHA,
Neuro-fuzzy approach.
473 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
1.INTRODUCTION
Image denoising also known as noise reduction is the process of eradicating the noise from
an image during edge preservation process. Normally, edge preserving process is one of the
image processing techniques that remove noise or textures while preserving sharp edges.
In these days, digital images play the most signicant role in dierent applications such as
magnetic satellite box, computer tomography, resonance imaging, and ecological systems.
In general, the images are captured by the sensors. In some cases, like faulty components,
data acquisition problems and normal phenomena interference, the most signicant data
can be degraded. As well, the noise may be occurred by the inaccurate compression and
transmission. As a result, image denoising process is needed as primary process in the image
processing to compensate the data corruption. Still, image denoising technique has lot of
challenges since the artifacts are occurred during noise removal which blurs the images.
Hosotani et al. (2015) introduced an image denoising process with edge preserving and
segmentation based on the Mask NHA (M- NHA). In this process, the zero-mean white
Gaussian noise was eliminated by using the high-resolution frequency analysis. The regions
including identical texture were analyzed on the noisy image. The non-uniform regions
obtained by segmentation process were analyzed by using M-NHA to improve the Peak
Signal-To-Noise Ratio (PSNR). Conversely, an optimization of the parameters used in
the segmentation was not ecient and the threshold value for edge detection was xed to
suppress the unwanted data. Hence, an OAMNHA based image denoising technique was
proposed by using SVM and rey optimization algorithms. In this technique, Support
Vector Machine (SVM) was applied for learning the parameters used in the segmentation
and rey algorithm was used for optimizing the threshold values for many noisy images.
However, the eciency of denoising was depending on the homogeneous texture
segmentation since it uses feature vectors in spatial domain that cannot be provided to search
similar patches from arbitrary regions. Therefore, the frequency domain coecients have
been used in image segmentation. Initially, the spatial domain image was transformed into
frequency domains such as Wavelet Transform (OAMNHA-WT), Contourlet Transform
(OAMNHA-CoT) and Curvelet Transform (OAMNHA-CT) separately. Then, the image
was decomposed into dierent subband images and homogeneous patches were explored
474 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
by using frequency coecients for achieving segmentation process. At last, 2D NHA and
inverse frequency transforms were applied independently for merging all subbands together.
Based on this technique, the noisy pixels are eciently removed via segmentation process.
Nonetheless, the region boundary was dicult to determine within a noisy image. The
segmentation of noisy images has fuzziness in the collection of the region margin. It was
assumed that fuzzy boundaries must be dened as edge regions. For this reason, canny edge
detector was used for detecting the edge positions with homogeneous textures. But, the time
consumption of this detection method was high due to its high computational complexity.
Therefore, in this article, an edge-preserving segmentation is performed based on the NF
approach. In this technique, the considered noisy images denoted in the frequency domain
transforms are applied to the NF edge detector to detect the edge regions and homogenous
textures from a noisy image. This edge segmentation and detection has the Neuro-fuzziness
in the choice of region boundary. Then, the edge regions are removed and the texture
regions are used for restoring the noise-free images by applying the OAMNHA technique.
Therefore, the boundary distortion is reduced by dening the texture boundaries due to
segmentation and detection process. Also, the accuracy of the edge- preserving segmentation
and detection is eciently increased.
The remaining part of the article is organized as follows: Section II presents dierent
image denoising approaches which are related to the proposed technique. Section III
claries the methodology of proposed image denoising technique. Section IV illustrates the
experimental results and Section V concludes the entire discussion.
Chen et al. (2013) proposed an edge preserving image denoising with a closed form solution.
In this method, a novel pixel-based algorithm was proposed that formulates the image
denoising problem as the Maximum-A-Posterior (MAP) estimation problem by using
Markov Random Fields (MRF). This was converted into a continuous label assignment
problem based on a Gaussian MRF model and then a closed form globally optimal solution
was obtained. The pre- estimated image edge information was added into the energy
function construction to preserve the image structures. Moreover, patch similarity based
pairwise interaction was performed to better preserve the image details and achieve high
robustness. However, Mean Squared Error (MSE) of this algorithm was high.
475 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
Jiao and Huang (2014) proposed a new wavelet packet transform adaptive threshold image
denoising method based on edge detection. By edge detection method, the wavelet packet
coecients corresponding to detected edge and other non-edge wavelet packet coecients
were treated by dierent threshold. By using the relativity among wavelet packet coecients
and neighbor dependency relation, the new variance neighbor estimate method was adopted
and then the adaptive threshold was generated. On the other hand, the computational
complexity of this method was high.
Gao, Wang and Liu (2015) proposed an image denoising method based on edge detection
and pre-thresholding wiener ltering of multi-wavelets fusion for removing Gaussian noise
from digital images. Initially, the noisy image was decomposed by using multiple wavelets
and the edge of image was detected through wavelet multi-scale edge detection. Based on
this, the wavelet coecients belonging to the edge position were dealt with the improved
wavelet threshold method whereas the others were dealt with the pre-thresholding wiener
ltering. At last, the fusion algorithm was used based on wavelet analysis for obtaining the
denoised image. However, the computational burden i.e., computation time was high and
it requires an adaptive threshold selection using advanced optimization methods to further
improve the image denoising performance.
Guo, Zhang and Zhang (2018) proposed an edge-preserved image denoising algorithm
based on local adaptive regularization which can adaptively adjust denoising in accordance
with dierent regions of noisy image by containing residual local energy function. In this
method, detailed information of image was also well preserved at the process of denoising.
However, this method has high time complexity. Yang et al. (2016) proposed a block
thresholding image denoising using Dual-Tree Complex Wavelet Transform (DTCWT) for
removing the additive white Gaussian noise from the images. Initially, DTCWT was applied
on noisy image for obtaining the complex coecients with properties of approximate shift
invariance and directional selection. Moreover, block thresholding scheme was adopted in
denoising process to select the optimal block size and threshold at each resolution level by
reducing Stein’s unbiased risk estimate. However, this method was suitable for preserving
the uniform regions only.
476 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
Jain and Tyagi (2017) proposed an adaptive edge-preserving image denoising technique
by using patch-based weighted-Singular Value Decomposition (SVD) ltering in wavelet
domain. In this ltering method, a group of non-local similar patches were considered for
each local patch in the subband and the weights were estimated for this patch group based
on the signicance of its singular values. After that, these weights were used as a threshold
for SVD ltering of the considered patch group. This process was applied to each local
patch and all patches were aggregated together for obtaining the thresholder results in the
subband. However, this algorithm was not able to obtain the depth of the image which has
the shadow eect and the intensity values ranged in the black to gray.
Zhao and Shang (2018) proposed an adaptive edge-detection method i.e., a method of
iterative threshold edge detection based on histogram. In this method, multi-scale wavelet
transform was used for preprocessing the image in which the image detains were highlighted
and the eect of manual setting lter coecients was avoided. The variation of gray values
between the pixels of local region was used for computing the gradients and the gradient
directions were extended to four directions. The adaptive method was used for computing
the threshold of edge-detection and the image was represented by histogram. After that,
the ratio of the number of pixels in the bar and the total numbers of pixels were used for
selecting the initial threshold.
The regions on both sides of the initial threshold were used for calculating the high
threshold and low threshold iteratively. The detection errors, connection errors and the
pseudo-edges caused by selecting the threshold articially were avoided. However, the
Signal-to-Noise Ratio (SNR) of this method was less. Guo et al. (2016) recommended a
novel edge-preserving image denoising algorithm which treats the low and high-frequency
components of the image separately. For restoration of high-frequency components,
a neighborhood regression method was proposed. Moreover, an energy minimization
function was developed for combining the low and high-frequency components composed
for obtaining the result as denoised. However, the average run time of this algorithm was
high and also the performance eciency was not eective.
477 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
2. MATERIALS AND METHODS
In this section, the proposed NF-OAMNHA-CT based image denoising technique using
NF approach is explained briey. Initially, the image i(x,y) with size M×N is transformed
into frequency domain such as CT. Then, the image is given to the NF edge detector to
detect the edge regions and homogenous textures as follows:
The structure of the proposed Neuro Fuzzy edge detector shown in Figure 1(a) has four
NF networks which are functioning as sub-detectors in the four directions, namely vertical,
horizontal, right diagonal and left diagonal, respectively. Each sub-detector can operate
on a window size of 3x3 which is shown in Figure 1(b). Also, each sub-detector estimates a
dierent neighbourhood correlation between the center pixel of the ltering window and
two of its neighbour’s.
Figure 1(a). Structure of the NF Edge Detector.
Source: own elaboration.
Each NF sub-detector is a rst-order Sugeno type Fuzzy Inference System (FIS) with
3-inputs and 1-output. Each input has 3 generalized bell type membership functions and
the output has a linear membership function (m). The input-output correlation of any of
the NF sub-detectors is as follows:
478 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
Consider i_1,i_2,i_3 are the inputs of the NF sub-detector and e is its output. Each possible
combination of inputs and their related membership functions is denoted by a rule in the
rule base of the NF sub-detector. As the NF sub-detector consists of 3 inputs and each input
has 3 membership functions, the rule base has 27 rules totally which are given below:
if (i1 is m11 ) and (i2 is m21 ) and (i3 is m31 ), then R1 = F1 (i1, i2, i3 )
if (i1 is m11 ) and (i2 is m21 ) and (i3 is m32 ), then R2 = F2 (i1, i2, i3 )
if (i1 is m11 ) and (i2 is m21 ) and (i3 is m33 ), then R3 = F3 (i1, i2, i3 )
if (i1 is m11 ) and (i2 is m22 ) and (i3 is m31 ), then R4 = F4 (i1, i2, i3 )
·
·
·
if (i1 is m13 ) and (i2 is m23 ) and (i3 is m33 ), then R27 = F27 (i1, i2, i3 )
Here, mij is the jth membership function of the ith input, Rk is the output of the kth rule
and Fk is the kth output membership function. The input membership functions (mij) are
generalized bell type which is dened as follows:
(1)
The output membership functions (Fk) are as:
Fk= dk0 + dk1i1+ dk2i2 + dk3i3 (2)
For i = j = 1,2,3 and k=1,2,…,27
The factors a, b, c and d in the above equations are constants that dierentiate the shape
of the membership functions. The optimal values of these parameters are determined by
learning process. Each NF sub-detector is learned individually. The setup used for learning
is shown in Figure 2.
Here, the parameters of the Neuro Fuzzy sub-detectors under learning are iteratively
ne-tuned so that its output converges to the output of the ideal edge detector which can
479 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
properly detect the positions of the edge pixels in the given image. The ideal edge detector
output represented by the target training image is used only for learning process.
Figure 2. Training of the Sub-detectors.
Source: own elaboration.
The output of the Neuro Fuzzy sub-detector is the weighted average of the individual rule
outputs. The weighting factor (wk) of each rule is computed by estimating the membership
expressions in the antecedent of the rule. This is achieved by converting the input values
to fuzzy membership values by using the input membership functions and applying and
operator to these membership values. Thus, the weighting factors of the rules are computed
as follows:
w1= m11 (i1).m21 (i2).m31 (i3)
w2 = m11 (i1).m21 (i2).m32 (i3)
w3 = m11 (i1).m21 (i2).m33 (i3)
·
·
·
w27 = m13 (i1).m23 (i2 ).m33 (i3 ) (3)
After computing the weighting factors, the output of the NF sub-detector can be obtained
by computing the weighted average of the individual rule outputs as:
480 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
(4)
The outputs of the Neuro Fuzzy sub-detectors values are given to the postprocessor to
generate the nal NF network output. Normally, the postprocessor is used for computing
the average value of the sub-detector outputs and comparing this value with a threshold.
The threshold value is half of the available dynamic range for the pixel luminance values.
For 8-bit images where the pixel values range between 0 and 255, the threshold value is
taken as 128. The input and output correlation of the postprocessor is described as follows:
Consider e1, e2, … ,e1 is the outputs of the NF sub-detectors where l is the number of NF
sub-detectors used. The output of the postprocessor is computed in two steps:
Step 1: The average value of the individual NF sub-detectors outputs is computed as:
(5)
Step 2: The computed eAV value is altered into 0 (black) or 255 (white) by relating it with
the threshold as:
(6)
In equation (6), e(x,y) represents the result of the postprocessor i.e., the output of the
Neuro Fuzzy edge detector that denotes the conclusion whether the center pixel of the
ltering window is an edge pixel or not. This is continued until all pixels of the noisy image
are classied properly. Thus, the edges from the noisy image are extracted eciently and
then the edge regions are dened accurately. After that, the texture boundaries formed
and OAMNHA technique is applied for each homogeneous texture region in order to
reconstruct the noise-free images with the increased segmentation accuracy and reduced
computational complexity.
481 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
3.RESULTS
In this section, the performance of proposed NF-OAMNHA-CT based image denoising
technique is evaluated and compared with the existing OAMNHA-CT technique using
MATLAB. In this experiment, target images with various features are gathered from LIVE
(The Laboratory for Image Video Engineering) image database at the University of Texas
at Austin. The comparison is obtainable in terms of PSNR, Mean Absolute Error (MAE)
and Structural Similarity Index (SSIM). The results of proposed NF-OAMNHA-CT based
image denoising technique are shown in Figure 3.
Figure 3. Results of Proposed and Existing Denoising Technique for Lena Image.
Source: own elaboration.
3.1. PSNR
It is the ratio of the maximum signal power to noise power and computed as:
(7)
(8)
The comparison of denoising PSNR which is made between proposed NF-OAMNHA-CT
and existing OAMNHA-CT techniques is shown in Table 1. It shows that the proposed
NF-OAMNHA-CT technique achieves higher PSNR compared with the OAMNHA-CT
482 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
technique for image denoising. For example, consider the statue with σ=5. Then, the PSNR
of proposed NF-OAMNHA-CT technique is 4.5% higher than OAMNHA-CT technique.
Table 1. Comparison of denoising PSNR results (σ: Noise density).
σCameraman Bikes
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 43.88 46.25 44.86 46.62
10 42.33 45.21 40.68 42.86
15 39.14 41.38 38.99 41.03
20 38.86 40.48 36.07 38.55
25 36.72 39.06 35.66 37.39
30 35.69 38.13 34.53 36.72
σOcean Statue
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 45.99 48.03 45.98 48.05
10 41.67 43.49 41.86 43.61
15 40.06 42.99 40.04 42.89
20 38.35 40.11 39.22 41.33
25 36.88 38.62 36.93 38.74
30 35.96 37.85 35.76 37.92
σBarbara Lena
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 45.91 47.68 46.32 48.86
10 42.03 44.21 42.98 44.37
15 39.81 41.19 41.05 43.22
20 38.94 40.06 39.86 42.31
25 36.97 38.34 39.17 41.95
30 35.59 37.95 37.99 39.04
σCoinsfountain Lighthouse
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 44.92 46.28 45.90 47.68
10 41.77 43.51 40.68 42.31
15 38.25 40.85 39.17 41.85
20 36.88 38.43 36.89 38.47
25 35.06 38.36 35.95 37.52
30 34.18 36.74 35.06 37.19
σStream
OAMNHA-CT NF-OAMNHA-CT
5 40.97 43.06
10 37.81 39.65
15 36.07 38.71
483 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
20 34.15 36.34
25 41.92 43.87
30 41.28 43.43
Source: own elaboration.
3.2. MAE
It is the measure used to calculating the average magnitude of error in the prediction groups
and computed as:
(9)
In equation (9), Pi is the prediction value, Yi is the true value and εi is the absolute error.
The comparison of denoising MAE for proposed NF-OAMNHA-CT and the existing
OAMNHA-CT techniques is shown in Table 2. It shows that the proposed NF-OAMNHA-
CT technique has less MAE than the OAMNHA-CT technique for image denoising. If the
statue with σ=5, then the MAE of proposed NF-OAMNHA-CT technique is 26.67% less
than OAMNHA-CT technique.
Table 2. Comparison of denoising MAE results.
σCameraman Bikes
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 1.65 1.12 1.70 1.24
10 2.55 1.96 3.65 3.18
15 3.44 2.91 4.81 4.36
20 4.29 3.73 6.06 5.49
25 5.51 5.05 7.46 6.93
30 5.68 5.17 8.58 8.01
σOcean Statue
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 1.83 1.35 1.95 1.43
10 2.99 2.47 2.96 2.46
15 3.92 3.39 3.87 3.33
20 4.28 3.71 4.52 4.09
25 5.08 4.56 5.20 4.75
30 5.63 5.12 5.74 5.27
σBarbara Lena
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 2.13 1.68 1.99 1.53
10 3.25 2.71 2.81 2.38
484 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
15 3.96 3.43 3.44 2.91
20 4.68 4.16 3.88 3.54
25 5.26 4.72 4.36 3.87
30 5.90 5.45 4.78 4.29
σCoinsfountain Lighthouse
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 2.54 2.08 2.25 1.73
10 3.96 3.42 3.31 2.86
15 5.18 4.67 4.28 3.74
20 6.24 5.71 4.98 4.51
25 7.07 6.64 5.60 5.49
30 7.89 7.53 6.05 5.52
σStream
OAMNHA-CT NF-OAMNHA-CT
5 3.27 2.76
10 5.52 5.04
15 7.34 6.88
20 8.74 8.25
25 9.98 9.41
30 10.90 10.53
Source: own elaboration.
3.3. SSIM
It is the similarity measure between any two images i(x,y) and computed as follows:
(10)
In equation (10), the averages of x and y are represented as μx and μy
. Also, the variances of
x and y are represented as σx
2 and σy
2 x,y respectively, c1 and c2 are constants. As well, the
covariance of x and y is denoted as σxy
.
The comparison of denoising SSIM for both proposed NF-OAMNHA-CT and the existing
OAMNHA-CT technique is shown in Table 3. It shows that the proposed NF-OAMNHA-
CT technique has higher SSIM compared to the OAMNHA-CT based image denoising
technique. For example, consider the statue with σ=5. Then, the SSIM of proposed NF-
OAMNHA-CT technique is 0.31% higher than OAMNHA-CT technique.
Table 3. Comparison of denoising SSIM results.
485 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
σCameraman Bikes
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 0.984 0.987 0.992 0.995
10 0.966 0.969 0.980 0.983
15 0.942 0.945 0.949 0.952
20 0.896 0.899 0.914 0.917
25 0.876 0.879 0.866 0.869
30 0.865 0.868 0.825 0.828
σOcean Statue
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 0.985 0.988 0.978 0.981
10 0.958 0.961 0.954 0.957
15 0.901 0.904 0.917 0.920
20 0.866 0.869 0.888 0.891
25 0.827 0.830 0.865 0.868
30 0.806 0.809 0.846 0.849
σBarbara Lena
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 0.985 0.988 0.968 0.971
10 0.966 0.969 0.943 0.946
15 0.953 0.956 0.927 0.930
20 0.939 0.942 0.910 0.913
25 0.917 0.920 0.896 0.899
30 0.911 0.914 0.884 0.887
σCoinsfountain Lighthouse
OAMNHA-CT NF-OAMNHA-CT OAMNHA-CT NF-OAMNHA-CT
5 0.986 0.989 0.976 0.979
10 0.956 0.959 0.947 0.950
15 0.921 0.924 0.927 0.930
20 0.874 0.877 0.880 0.883
25 0.826 0.829 0.867 0.870
30 0.794 0.797 0.838 0.841
σStream
OAMNHA-CT NF-OAMNHA-CT
5 0.993 0.996
10 0.972 0.975
15 0.913 0.916
20 0.842 0.845
25 0.778 0.781
30 0.722 0.725
Source: own elaboration.
486 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
4. CONCLUSIONS
In this paper, NF-based edge-preserving segmentation is proposed to enhance the
OAMNHA-based image denoising technique in dierent frequency transform domains.
Initially, the considered image is fed into the Neuro Fuzzy edge detector to classify the pixel
of the given image is whether edge pixel or not. For all pixels in the image, Neuro Fuzzy
ltering process is repetitive and the edges are extracted precisely from the noisy image.
Then, the extracted edges are dened as the edge regions. Also, the texture boundaries
are constructed based on the homogeneous texture segmentation. For each segment,
OAMNHA technique is applied to restore the noiseless image eciently. Therefore, the time
consumption for edge detection is reduced and the accuracy of the segmentation process is
also increased. Finally, the experimental results proved that the proposed NF-OAMNHA-
CT technique achieves higher PSNR, SSIM and less MAE than the OAMNHA-CT based
image denoising technique. In many applications, e.g., medical or satellite imaging, the
edges are key features and thus must be preserved sharp and undistorted in smoothing/
denoising. Also, edge preserving denoising is useful for the images capturing from camera
which is an optical device and prone to sensor noise, especially in dark environments or
environments with extreme high dynamic range.
ACKNOWLEDGEMENT
With pride and pleasure, I use this opportunity to place on record a deep sense of gratitude
to my guiding light Dr. ANNA SARO VIJENDRAN, M.C.A., M. Phil., Ph. D., Dean,
School of Computing, Sri Ramakrishna College of Arts & Science (Autonomous), for her
guidance and help throughout this research work and in writing this research paper. I deeply
express my cordial thanks to all the Faculty Members of Department of Computer Science
and Librarians of Sri Ramakrishna College of Arts & Science (Autonomous).
REFERENCES
Chen, S., Liu, M., Zhang, W., & Liu, J. (2013). Edge preserving image denoising with a
closed form solution. Pattern Recognition, 46(3), 976-988. https://doi.org/10.1016/j.
patcog.2012.08.014
487 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
Gao, L., Wang, G., & Liu, J. (2015). Image denoising based on edge detection and
prethresholding Wiener ltering of multi-wavelets fusion. International Journal of
Wavelets, Multiresolution and Information Processing, 13(05), 1550031. https://doi.
org/10.1142/S0219691315500319
Guo, F., Zhang, C., & Zhang, M. (2018). Edge-preserving image denoising. IET Image
Processing, 12(8), 1394-1401. https://doi.org/10.1049/iet-ipr.2017.0880
Guo, L., Chen, W., Liao, Y., Liao, H., & Li, J. (2016). An edge preserved image denoising
algorithm based on local adaptive regularization. Journal of Sensors, 2016. Article ID
2019569. http://dx.doi.org/10.1155/2016/2019569
Hosotani, F., Inuzuka, Y., Hasegawa, M., Hirobayashi, S., & Misawa, T. (2015).
Image Denoising With Edge-Preserving and Segmentation Based on Mask NHA.
IEEE Transactions on Image Processing, 24(12), 6025-6033. https://doi.org/10.1109/
TIP.2015.2494461
Indupriya, K., & Kumar, P. R. (2017). An Optimum Adaptive Parameterized Mask
NHA Based Image Denoising. International Journal of Engineering and Technology (IJET),
9(3), 2389-2399. https://doi.org/10.21817/ijet/2017/v9i3/1709030190
Jain, P., & Tyagi, V. (2017). An adaptive edge-preserving image denoising technique
using patch-based weighted-SVD ltering in wavelet domain. Multimedia Tools and
Applications, 76(2), 1659-1679. https://doi.org/10.1007/s11042-015-3154-8
Jiao, Y., & Huang, B. W. (2014). A new adaptive threshold image denoising method
based on edge detection. Applied Mechanics and Materials, 678, 137-142. https://doi.
org/10.4028/www.scientic.net/AMM.678.137
Vijendran, A. S., & Kumarasamy, I. (2018). Image Denoising by Optimum Adaptive
Parameterized Mask NHA in Various Frequency Domains. International Journal
of Pure and Applied Mathematics, 119(18), 3405-3413. https://www.researchgate.
net/publication/327765365_Image_Denoising_by_Optimum_Adaptive_
Parameterized_Mask_NHA_in_Various_Frequency_Domains
488 https://doi.org/10.17993/3ctecno.2021.specialissue8.471-489
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue
Noviembre 2021
Yang, P., Yang, G., Gai, S., & Chen, J. (2016). Block thresholding image denoising with
dual-tree complex wavelet transform. Journal of Intelligent & Fuzzy Systems, 30(6),
3197-3207. https://doi.org/10.3233/IFS-152063
Zhao, H., & Shang, M. (2018). An adaptive edge-detection method based on
histogram. Modern Physics Letters B, 32(34n36), 1840088. https://doi.org/10.1142/
S0217984918400882
