An Efficient Image Denoising Approach for the Recovery of Impulse Noise

Bulletin of Electrical Engineering and Informatics
ISSN: 2302-9285
Vol. 6, No. 3, September 2017, pp. 281~286, DOI: 10.11591/eei.v6i3.680  281
Received June 3, 2017; Revised August 14, 2017; Accepted August 28, 2017
An Efficient Image Denoising Approach for the
Recovery of Impulse Noise
S. Rajkumar*
1
, G. Malathi
2
School of Computing Science and Engineering, VIT University, Chennai, India
*Corresponding author, e-mail: rajkumar.srinivasan@vit.ac.in
1
, malathi.g@vit.ac.in
2
Abstract
Image noise is one of the key issues in image processing applications today. The noise will affect
the quality of the image and thus degrades the actual information of the image. Visual quality is the
prerequisite for many imagery applications such as remote sensing. In recent years, the significance of
noise assessment and the recovery of noisy images are increasing. The impulse noise is characterized by
replacing a portion of an image’s pixel values with random values Such noise can be introduced due to
transmission errors. Accordingly, this paper focuses on the effect of visual quality of the image due to
impulse noise during the transmission of images. In this paper, a hybrid statistical noise suppression
technique has been developed for improving the quality of the impulse noisy color images. We further
proved the performance of the proposed image enhancement scheme using the advanced performance
metrics.
Keywords: image quality, impulse noise, visual quality, image denoising, image sharpen, image
smoothing
1. Introduction
Noise is generally accumulated in to images during acquisition, transmission and
compression. Image denoising is perhaps the most powerful operation of image processing and
computer vision. The impulse noise is occurred due to the fault encountered in camera during
the time of image transmission and reception, external disturbances in a booming natural
environment, electromagnetic interferences, and timing problems in analog to digital conversion.
The impulse noise is also called as spike noise, random noise or independent noise. Since the
independent pixels in the image bears no relationship with the neighboring pixels.
The impulse noise is similar to salt and pepper noise with a slight difference. The
impulse noise is characterized by random occurrences of bright and dark pixels and the impulse
noise is characterized by random occurrences of bright dots. Mostly, the impulse noise is
occurred in the electronic devices that transmits or receives the signal. The impulse noise arises
from sudden spikes, which sometimes produces the annoying spots in the television signal.
The dust particles can also cause this type of noise. Uncertainties are the major impact
of impulse noise. This fact makes image de-noising a difficult task. Understanding the
uncertainties can enhance the quality of image de-noising. The algorithm in [1] uses two stage
filtering algorithm in which the median value is computed twice and is used to remove the high
density corrupted images. Amarjit Roy et.al. [2] proposed an artificial neural network (ANN)
based fuzzy filter for removing the impulse burst. The ANN is used to classify the noise and
signal from the image and based on the classification the fuzzy filtering is applied to adjust the
noise and signal. In [3], an efficient method for detecting the impulse noise was developed. In
this approach, once the noisy pixel is detected, the subsequent noise can be removed by
various existing techniques such as switching median filters. The experimental result in [4]
examines the contrast enhancement and tone mapping algorithms which applies to satellite
images. In [5], an Additive White Gaussian Noise (AWGN) removal method was using the high
resolution frequency concepts such as Non Harmonic Analysis (NHA) and is yielding good
results in numerous applications such as image tomography and motion analysis. Enming Luo
et. al [6] developed expectation-maximization (EM) adaptation method to denoise the document
images. The above method yields good results when compared to existing denoising algorithm
such as Expected Patch Log Likelihood (EPLL). The traditional noise removal methods cannot

 ISSN: 2302-9285
Bulletin of EEI Vol. 6, No. 3, September 2017 : 281 – 286
282
adequately remove the noise from the corrupted images. For this reason, Pok et. al [7], has
been developed a separate class of nonlinear filters as based on homogeneous information for
the removal of impulse noise. Tadahiro Azetsu et. al [8], use the rank order statistics filter to
improve the quality of image which is corrupted due to impulse noise and is compared with
traditional methods. The model in [9], is used to remove the Charge-Coupled device (CCD)
noise and the experimental results proves the edge information as well. In [10], an optimal
contrast enhancement algorithm is developed and is used to improve the contrast of the fog
images. Bhat et.al in [11], is developed a hybrid approach that combines the back propagation
neural network and adaptive median filtering in order to remove the salt and pepper noise and
improve the image quality. In [12], a combined framework has been derived to restore the
quality astronomy images. The above framework use the phase preserving technique,
normalization method and Lucy debluring algorithm. A regularized minimization algorithm has
illustrated by Fahmy [13], is used to minimize the error between noisy image and filtered image.
Gonzalez-Hidalgo1 et.al [14], proposed a hybrid approach that inturn contain 2 stages. In the
first stage, an impulse noise detector based on the fuzzy tophat transforms is used to detect the
pixels which are likely to be composed by impulse noise. In the second level, the image is
filtered using a regularized method that uses fuzzy open-close or fuzzy close-open sequences
and is applied only to those selected infected pixels.
The technique in [15] aims to identifies the noisy pixels and replace the noisy pixels by
the sparsity information. Xiao Kang et.al [16], presents a Novel Adaptive Switching Median
(NASM) filter that effectively handles the salt and pepper noisy image and preserves the image
details as well. Jinli Suo et al. can [17] remove highlight from high resolution images and is more
complex compared to other existing approaches. A blind image restoration method in [18], is
used to compute the noise variance and blur estimation without any prior knowledge. In [19], a
novel Fuzzy Impulse noise Detection and Reduction Method (FIDRM) is proposed. The above
filter has two phases: The first phase is the detection phase and the filtering phase. The first
phase incorporates the fuzzy rules to identify whether the pixel is corrupted with impulse noise
or not. The second phase is used for removing the impulse noise.
An efficient novel approach [20] was developed for de-focused images. The de-focus is
occurred due to improper focus of camera lens. The model [20] can produce visually appealing
HDR (High Dynamic Range) image. Two different exposure images are used as the inputs in
the model. The shaken pixels between the images are identified. The blur pixels are removed
and optimal quality image is obtained by integrating the large and small exposure images. In
[21], the image quality is assessed by extracting the partial information from the image. The
partial information is taken from the input image at the source point and the received image at
the recipient side. An efficient quality assessment system was proposed and that provides more
accurate and user friendly. In this method, certain features of the source image were taken
through steerable pyramid decomposition algorithm and the image quality is measured by using
the Reduced Reference (RR) method.
This paper has been organized in the following manner:
Section I deals with the introduction of the paper. Section 2 explains the motivation of
the work. The research method-novel Hybrid Impulse Noise Removal (HINR) filter is illustrated
in Section 3. The Section 4 discusses the results and analysis. Section 5 explains the
conclusion and finally the future enhancement of our research method is explained in Section 6.
2. Motivation of the Work
Apart from the several techniques in image processing, the image denoising process is
a rapidly evolving field with growing applications. Removal of noise from a signal is an important
problem in image and signal processing. Noise is a practical problem raised in many systems
such as satellite, medical and seismology.
The noise filtering is the form of image denoising and it becomes necessary to remove
the noise and unwanted distortions such as impulse noise and mixed noise. Because of their
wide applications, denoising is of great importance in signal and image processing. Since much
effort was not taken, an initiative has been made to perform research in this area.

Bulletin of EEI ISSN: 2302-9285 
An Efficient Image Denoising Approach for the Recovery of Impulse Noise (S. Rajkumar)
283
3. Research Method- Hybrid Impulse Noise Removal (HINR) Filter
Earlier, the Order Statics Filters and Rank Order Rank Selection Filters were used to
remove the impulse noise. But, those filters does not able to distinguish between minute details
such as sharp edges and noise, also its performance becomes poor with increasing level of
noise density. To eliminate all these drawbacks, in this paper we propose a Hybrid Impulse
Noise Removal (HINR) filter. This filter is used to identify and handle high density impulse noisy
color images and it covers wide range of samples of an image. It also preserves the fine details
such as edges as well.
The novel HINR filter algorithm comprised of the following steps:
Step 1:
Convert the RGB input image in to grayscale using the relation:
Grayscale=R*0.21 + G*0.72 + B*0.01 (1)
Step 2:
Let n be the size of the window. Let us consider n=9. Since, the entire image is divided
in to 3X3 regions. Consider f(x,y) is the center pixel. With a zero padding process, all the pixels
are covered in this proposed algorithm. The mask 3X3 is traversed across the entire image start
from (0,0) co-ordinates to the entire size of the image.
a. if f(x,y)<255, then the 8 neighborhoods of the center pixel will be retained.
b. If f(x,y)==255, then the corresponding mask is subjected to the following conditions:
Condition-1:
If (f(x-1,y+1)==255) and if (f(x,y+1)==255) and if (f(x+1,y+1)==255) and if (f(x-1,y)==255)
and if (f(x+1,y)==255) and if (f(x-1,y-1)==255) and if (f(x,y-1)==255) and if (f(x+1,y-1)==255)
then
Compute the sum using the following relation:
Sum=f(x-1,y+1)+f(x,y+1)+f(x+1,y+1)+f(x-1,y)+f(x+1,y)+f(x-1,y-1)+f(x,y-1)+f(x+1,y-1) then the
average is to be computed and is replaced in the center pixel.
Condition-2:
Initiate Sum=0
If (f(x-1,y+1) !=255) then
Sum=Sum+f(x-1,y+1)
If (f(x,y+1)!=255) then
Sum=Sum+f(x,y+1)
If (f(x+1,y+1)!=255) then
Sum=Sum+f(x+1,y+1)
If (f(x-1,y)!=255) then
Sum=Sum+f(x-1,y)
If (f(x+1,y)!=255) then
Sum=Sum+f(x+1,y)
If (f(x-1,y-1)!=255) then
Sum=Sum+f(x-1,y-1)
If (f(x,y-1)!=255) then
Sum=Sum+f(x,y-1)
If (f(x+1,y-1)!=255) then
Sum=Sum+f(x+1,y-1) then the average is to be computed and is replaced in the center
pixel.
Step 3:
The output image is obtained from step (2) is the pure filtered image and thus the
impulse noise is eliminated. But, the output image is imposed by the smoothing effect. Further,
to avoid this smoothing effect, a sharpening filter is used.
The image obtained from step (2) is blurred (using box blur technique).
Step 4:
Subtract the blurred image from the obtained image through step (2). The resultant
image contains only the enhanced edge information.
Step 5:
The image obtained from step (4) is added to the image obtained from step (2) to get
the sharpened image and thus the smoothing effect in the filtered image is eliminated.

 ISSN: 2302-9285
284
4. Results and Analysis
The Image denoising process is vital part of many image and signal processing
applications. In real time applications, the images are corrupted due to atmospheric turbulence,
camera motion, aperture effects of the camera. Therefore an efficient image filter is required to
further processing or analyzing the images. In our paper, a color image with impulse noise is
considered as input and is shown in Fig.1. The size of the image is 256X256. As the
preprocessing step, the input is image is converted in to grey scale and is shown in Fig.2. The
results using existing fuzzy filter as explained in chapter 1 is shown in Fig.3. The filtered image
with our proposed method is shown in Fig.4. The performance of our proposed method is
assessed by using MSE, PSNR and structural content metrics. The results of the above
measures are shown in Table 1.
Figure 1. Impulse noisy Lena color image Figure 2. Impulse noisy grey scale image
Figure 3. Result of Fuzzy Filtering Technique Figure 4. Result of our proposed HINR method
Table 1. MSE, PSNR and Structural Content Comparison for Colored Lena Image with
Fuzzy Filter
Filters MSE PSNR(db)
Structural
Content
Fuzzy filter 718.82 21.95 0.37
Proposed
HINR
319.56 24 0.93
The Mean Square Error (MSE) metric is the dispersion metric and it can be used to
quantify the image enhancement algorithm in which it is applied to removal of noise and blur. If
the MSE value increases, then the image degradation increases. When MSE value reaches
zero then pixel by pixel matching of images becomes perfect.
The MSE is calculated by using the following relation:
MSE= ∑ ∑ ) )) (2)

Bulletin of EEI ISSN: 2302-9285 
An Efficient Image Denoising Approach for the Recovery of Impulse Noise (S. Rajkumar)
285
The Peak Signal to Noise Ratio (PSNR) is the most significant metric which is used to
measure the visual quality of the restored image when it is corrupted due to noise and blur.
Higher the value of PSNR, indicates higher the quality rate. The PSNR value is computed by
using the following expression:
PSNR=10log10 (3)
The Structural Content (SC) deals with spatial arrangements of pixels in an image.
This metric brings out the similarity between two images.
SC can be given as:
Structural Content=
∑ ∑ ) )
∑ ∑ ))
(4)
where M is the number of pixels in horizontal direction, N is the number of pixels in vertical
direction, x(i,j) is the noisy image at i and j co-ordinates and y(i,j) is the filtered image at i and j
co-ordinates.
5. Conclusion
The impulse noise is manifested by the random occurrences of pixel values. The impact
of the impulse noise is the serious issue in image enhancement field. Therefore an effective,
intuitive and robust denoising scheme is required to eliminate the impulse burst present in the
real time images. Hence, in this paper, we proposed a filtering technique for removing the
impulse noise in a color image. The results obtained using the proposed filter technique ensures
noise free and quality of the image. It is shown that this filtering technique can provide an
effective removal of impulse bursts. It preserves the edges of the image and covers wide range
of samples. The quantitative analysis also proves that our HINR outperforms the existing Fuzzy
filter. So, we can conclude that the resulting technique is a versatile and provides greatly
improved efficiency while maintaining image quality and robustness.
6. Future Enhancement
The work can be extended to denoise any type of degradations such as Gaussian
noise, Salt and pepper noise, Laplacian noise etc. This approach is to be modified and applied
to the noises which affect the quality of satellite image
Acknowledgement
We wish to express our sincere gratitude and thanks to the reviewers for their
constructive comments and valuable suggestions and we also highly pleased to our university
for providing such a platform to carry out this research paper.
References
[1] Vikas Gupta, Rajesh Mahle, Umesh Gour. Image Noise Removal with Detail Preservation for Random
Valued Impulse Noise. IEEE Int.Conf.on Engineering. 2013: 1-5.
[2] Amarjit Roy, Salam Shuleenda Devi, RH Laskar. Impulse Noise Removal from Gray Scale Images
Based on ANN Classification Based Fuzzy Filter. In Proc. of IEEE Int. Conf. on Computational
Intelligence and Networks. September 2016: 97-101.
[3] Isha Singh, Om Prakash Verma. High Density Impulse Noise Detection using Fuzzy C-means
Algorithm. Defence Science Journal. January 2016; 66(1): 30-36.
[4] Jose-Luis Lisani, Julien Michel, Jean-Michel Morel, Ana Belén Petro, Catalina Sbert. An Inquiry on
Contrast Enhancement Methods for Satellite Images. IEEE Trans. Geoscience and Remote Sensing,
(This article has been accepted for inclusion in a future issue of this journal. Content is final as
presented, with the exception of pagination) 2016.
[5] Fumitaka Hosotani, Yuya Inuzuka, Masaya Hasegawa, Shigeki Hirobayashi, Tadanobu Misawa.
Image Denoising with Edge-Preserving and Segmentation Based on Mask NHA. December 2015;
24(12): 6025-6033.

 ISSN: 2302-9285
286
[6] Enming Luo, Student Member, IEEE, Stanley H. Chan, Member, IEEE, and Truong Q. Nguyen.
Adaptive Image Denoising by Mixture Adaptation. IEEE Trans. Image Process. October 2016; 25(10):
4489-4503.
[7] G Pok, J Liu, AS. Nair. Selective Removal of Impulse Noise Based on Homogeneity Level Information.
IEEE Trans. Image Process. January 2003; 12(1): 85-92.
[8] Tadahiro Azetsu, Noriaki Suetake, Eiji Uchino. Trilateral Filter Using Rank Order Information of Pixel
Value for Mixed Gaussian and Impulsive Noise Removal. In Proc. of Int. Conf. IEEE 2013: 303–306.
[9] Hilda Faraji, W. James MacLean. CCD Noise Removal in Digital Images. IEEE Trans. Image Process.
September 2006; 15(9): 2676-2685.
[10] Liu Bo, Xiong Qingguo. Inland River Image Defogging Based on Optimized Contrast Enhancement. In
Proc. of IEEE Int. Conf. on Information Technology, Networking, Electronic and Automation Control
Conference. Sep 2016: 145-150.
[11] Bhat Jasra, Aniqa Yaqoob, Sanjay Kumar Dubey. Removal of High Density Salt and Pepper Noise
Using BPANN-Modified Median Filter Technique. In 6th IEEE Int. Conf. on Cloud System and Big
Data Engineering. July 2016: 78-82.
[12] Z Al-Ameen, D Mohamad, M Shafry MR, G Sulong. Restoring Degraded Astronomy Images using a
Combination of Denoising and Deblurring Techniques. International Journal of Signal Processing,
Image Processing and Pattern Recognition (IJSIP). Mar 2012; 5(1): 1-12.
[13] Fahmy MF. A New Total Variation Based Image Denoising and Deblurring Technique. IEEE
EUROCON. July. 2013; 21(4): 1669-1675.
[14] M González-Hidalgo1 S Massanet1 A Mir1 D. Ruiz-Aguileral. High-density Impulse Noise Removal
Using Fuzzy Mathematical Morphology. In Proc. of Int. Conf. of the European Society for Fuzzy Logic
and Technology (EUSFLAT 2013): 728-735.
[15] S Huang, J Zhu. Removal of Salt-and-pepper Noise Based on Compressed Sensing. IEEE Trans.
Electronic Letters. August 2010; 46(17): 1198-1199.
[16] Xiao Kang, Wei Zhu, KeJie Li, Jing Jiang. A Novel Adaptive Switching Median Filter for Laser Image
Based on Local Salt and Pepper Noise Density. Proc. of Int. Conf. IEEE 2011: 38-41.
[17] Jinli Suo, Dongsheng An, Xiangyang Ji, Haoqian Wang, Qionghai Dai. Fast and High Quality Highlight
Removal from A Single Image. IEEE Trans. Image Process. August 2016.
[18] Chong Yi, Tetsuya Shimamura. A Blind Image Deconvolution Method Based on Noise Variance
Estimation and Blur Type Reorganization. In Proc. of IEEE Int. Symposium on Intelligent Signal
Processing and Communication Systems (lSPACS). December 2011.
[19] Stefan Schulte, Mike Nachtegael, Val´erie De Witte, Dietrich Van der Weken, Etienne E Kerre. A
Fuzzy Impulse Noise Detection and Reduction Method. IEEE Trans. Image Process. May 2006; 15(5).
[20] S Rajkumar, G Malathi. A Novel approach for the recovery of de-focused color Images. International
Journal on Recent Researches in Science, Engineering and Technology (IJRRSET). May 2015; 3(5).
[21] S Rajkumar. An Advanced Technique for the Analysis of Image Quality Assessment for Distorted
Images in the Field of Image Processing Application. In Proc. of conference in Karunya University,
Coimbatore, India, Mar.2008: 84-88.

An Efficient Image Denoising Approach for the Recovery of Impulse Noise

More Related Content

What's hot (18)

Similar to An Efficient Image Denoising Approach for the Recovery of Impulse Noise (20)

More from journalBEEI (20)

Recently uploaded (20)

An Efficient Image Denoising Approach for the Recovery of Impulse Noise