The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.

3, June 2014
DOI : 10.5121/ijma.2014.6303 35

PERFORMANCE ANALYSIS OF IMAGE
DENOISING WITH WAVELET THRESHOLDING
METHODS FOR DIFFERENT LEVELS OF
DECOMPOSITION

Anutam
1
and Rajni
2

1
Research Scholar SBSSTC, Ferozepur, Punjab
2
Associate Professor SBSSTC, Ferozepur, Punjab

ABSTRACT

Image Denoising is an important part of diverse image processing and computer vision problems. The
important property of a good image denoising model is that it should completely remove noise as far as
possible as well as preserve edges. One of the most powerful and perspective approaches in this area is
image denoising using discrete wavelet transform (DWT). In this paper, comparison of various Wavelets at
different decomposition levels has been done. As number of levels increased, Peak Signal to Noise Ratio
(PSNR) of image gets decreased whereas Mean Absolute Error (MAE) and Mean Square Error (MSE) get
increased . A comparison of filters and various wavelet based methods has also been carried out to denoise
the image. The simulation results reveal that wavelet based Bayes shrinkage method outperforms other
methods.

KEYWORDS

Denoising, Filters, Wavelet Transform, Wavelet Thresholding


1. INTRODUCTION

Applications of digital world such as Digital cameras, Magnetic Resonance Imaging (MRI),
Satellite Television and Geographical Information System (GIS) has increased the use of digital
images. Generally, data sets collected by image sensors are contaminated by noise. Imperfect
instruments, problems with data acquisition process, and interfering natural phenomena can all
corrupt the data of interest [1]. Various types of noise present in image are Gaussian noise, Salt &
Pepper noise and Speckle noise. Image denoising techniques are used to prevent these types of
noises while retaining the important signal features [2]. Spatial filters like mean and median filter
are used to remove the noise from image. But the disadvantage of spatial filters is that these filters
not only smooth the data to reduce noise but also blur edges in image. Therefore, Wavelet
Transform is used to preserve the edges of image [3]. It is a powerful tool of signal or image
processing for its multi-resolution possibilities.

This paper is organized as follows: Section 2 presents types of noise. Section 3 presents Filtering
techniques. Section 4 discusses Wavelet based denoising techniques and various thresholding
methods. Finally, simulated results and conclusions are presented in Section 5 and 6 respectively.

The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
36
2 .TYPES OF NOISE
Various types of noise have their own characteristics and are inherent in images in different ways.
2.1. Amplifier Noise (Gaussian Noise)
The standard model of amplifier noise is additive, Gaussian, which is independent at each pixel
and independent of the signal intensity. In color cameras, blue colour channels are more amplified
than red or green channel, therefore, blue channel generates more noise [4].
2.2. Impulsive Noise
Impulsive noise is also called as salt-and- pepper noise or spike noise. This kind of noise is
usually seen on images. It consists of white and black pixels. An image containing salt and pepper
noise consists of two regions i.e. bright and dark regions. Bright regions consist of dark pixels
whereas dark regions consist of bright pixels. Transmitted bit errors, analog-to-digital converter
errors and dead pixels contain this type of noise [5].
2.3. Speckle Noise
Speckle noise is a multiplicative noise. It is a granular noise that commonly exists in and the
active radar and synthetic aperture radar (SAR) images. Speckle noise increases the mean grey
level of a local area. It is causing difficulties for image analysis in SAR images .It is mainly due
to coherent processing of backscattered signals from multiple distributed targets [4].
3. FILTERING TECHNIQUES
The filters that are used for removing noise are Mean filter and Median filter.
3.1. Mean Filter
The advantage of using this filter is that it provides smoothness to an image by reducing the
intensity variations between the adjacent pixels [6]. Mean filter is essentially an averaging filter.
It applies mask over each pixel in signal. Therefore, to make a single pixel each of the
components of pixel which falls under the mask are average filter. The main disadvantage of
Mean filter is that it cannot preserve edges.
3.2. Median Filter
One type of non linear filter is Median filter. By firstly finding the median value and then
replacing each entry in the window with the pixels median value, median filtering is done [7].
Median is just the middle value after all the entries made in window are sorted numerically, if
window has an odd number of entries. There is more than one median when window has an even
number of entries. It is a robust filter. To provide smoothness in image processing and time series
processing, median filters are used.
4. WAVELET TRANSFORM
Wavelet domain is advantageous because DWT make the signal energy concentrate in a small
number of coefficients, hence, the DWT of a noisy image consists of number of coefficients
having high Signal to Noise Ratio(SNR) while relatively large number of coefficients is having
low SNR. After removing the coefficients with low SNR, the image is reconstructed using inverse
DWT [3]. Time and frequency localization is simultaneously provided by Wavelet transform.
Moreover, wavelet methods represent such signals much more efficiently than either the original
domain or fourier transform [8].
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
37
The DWT is same as hierarchical sub band system where the sub bands are logarithmically
spaced in frequency and represent octave-band decomposition. Image is decomposed into four
sub-bands and critically sampled by applying DWT as shown in Fig. 1(a). These sub bands are
formed by separable applications of horizontal and vertical filters. Sub-bands with label LH1,
HL1 and HH1 correspond to finest scale coefficient while sub-band LL1 represent coarse level
coefficients [9] [3]. The LL1 sub band is further decomposed and critically sampled to find out
the next coarse level of wavelet coefficients as shown in Fig. 1(b). It results in two level wavelet
decomposition.

(a ) One- Level (b) Two- Level

Figure1. Image Decomposition by using DWT
4.1 Wavelet Based Thresholding
Wavelet thresholding is a signal estimation technique that exploits the capabilities of Wavelet
transform for signal denoising. It removes noise by killing coefficients that are irrelevant relative
to some threshold [9] .Several studies are there on thresholding the Wavelet coefficients. The
process, commonly called Wavelet Shrinkage, consists of following main stages:

Figure2. Block diagram of Image denoising using Wavelet Transform
Read the noisy image as input
Perform DWT of noisy image and obtain Wavelet coefficients
Estimate noise variance from noisy image
Calculate threshold value using various threshold selection rules or shrinkage rules
Apply soft or hard thresholding function to noisy coefficients
Perform the inverse DWT to reconstruct the denoised image.
4.1.1. Thresholding Method
Hard and soft thresholding techniques are used for purpose of image denoising. Keep and kill rule
which is not only instinctively appealing but also introduces artifacts in the recovered images is
the basis of hard thresholding [10] whereas shrink and kill rule which shrinks the coefficients
above the threshold in absolute value is the basis of soft thresholding [11]. As soft thresholding
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
gives more visually pleasant image and reduces the abrupt sharp changes t
thresholding, therefore soft thresholding is pref
The Hard Thresholding operator

D (U, ) =U for all |U|>
= 0 otherwise
The Soft Thresholding operation

D (U, ) = sgn(U)* max(0,|U| -
(a) Hard Thresholding (b)
4.1.2. Threshold Selection Rules
In image denoising applications,
selected [9]. Finding an optimal value for thresholding is not an easy task.
threshold then it will pass all the noisy coefficients and
but larger threshold makes more number of coefficients to zero, which
image and image processing may cause blur and artifacts, and hence the resultant
lose some signal values [16].
4.1.2.1. Universal Threshold
where

being the noise variance

asymptotic sense and minimizes the cost fu
assumed that if number of samples is large, then the universal threshold may give better estimate
for soft threshold [18].
4.1.2.2. Visu Shrink
Visu Shrink was introduced by Donoho
shrinkage is that neither speckle noise can be removed nor MSE can be minimized
deal with additive noise [20]. Threshold T can

Where is calculated as mean of absolute difference (MAD) which is a robust estimator
represents the size of original image.
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
gives more visually pleasant image and reduces the abrupt sharp changes that occurs in hard
thresholding, therefore soft thresholding is preferred over hard thresholding [12] [13]
[14] is defined as,


on the other hand is defined as ,
)

Hard Thresholding (b) Soft Thresholding [15]
Figure 3. Thresholding Methods
Threshold Selection Rules
In image denoising applications, PSNR needs to be maximized , hence optimal value should be
]. Finding an optimal value for thresholding is not an easy task. If we select a
will pass all the noisy coefficients and hence resultant images may
threshold makes more number of coefficients to zero, which provides smooth
image and image processing may cause blur and artifacts, and hence the resultant
T 2logM

being the noise variance and M is the number of pixels [17] .It is optimal threshold in
asymptotic sense and minimizes the cost function of difference between the function.
assumed that if number of samples is large, then the universal threshold may give better estimate
Visu Shrink was introduced by Donoho [19]. It follows hard threshold rule. The drawback
is that neither speckle noise can be removed nor MSE can be minimized
Threshold T can be calculated using the formulae [21],


is calculated as mean of absolute difference (MAD) which is a robust estimator
represents the size of original image.
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
38
hat occurs in hard
].

(1)
(2)
hence optimal value should be
If we select a smaller
may still be noisy
smoothness in
image and image processing may cause blur and artifacts, and hence the resultant images may
(3)
It is optimal threshold in
of difference between the function. It is
assumed that if number of samples is large, then the universal threshold may give better estimate
follows hard threshold rule. The drawback of this
is that neither speckle noise can be removed nor MSE can be minimized .It can only
,
(4)
(5)
is calculated as mean of absolute difference (MAD) which is a robust estimator and N
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
39
4.1.2.3. Bayes Shrink
The Bayes Shrink method has been attracting attention recently as an algorithm for setting
different thresholds for every sub band. Here sub-bands refer to frequency bands that are different
from each other in level and direction [22]. Bayes Shrink uses soft thresholding. The purpose of
this method is to estimate a threshold value that minimizes the Bayesian risk assuming
Generalized Gaussian Distribution (GGD) prior [13]. Bayes threshold is defined as [23],

(6)

Where

is the noise variance and

is signal variance without noise.

From the definition of additive noise we have,

w (x, y) = s(x, y)+n(x, y) (7)

Since the noise and the signal are independent of each other, it can be stated that ,

(8)

can be computed as shown below:

(x, y)

,
(9)

The variance of the signal,

is computed as

max(

2

2
, 0) (10)
5. SIMULATION RESULTS
Simulated results have been carried on Cameraman image by adding two types of noise such as
Gaussian noise and Speckle noise. The level of noise variance has also been varied after selecting
the type of noise. Denoising is done using two filters Mean filter and Median filter and three
Wavelet based methods i.e. Universal threshold, Visu shrink and Bayes shrink. Results are shown
through comparison among them. Comparison is being made on basis of some evaluated
parameters. Also the comparison of wavelet thresholding methods at different decomposition
level has been discussed. The parameters are Peak Signal to noise Ratio (PSNR), Mean Square
Error (MSE) and Mean Absolute Error (MAE).
PSNR = 10 log

db (11)

MSE =

(x, y)

(X(i, j)

(, ))

(12)

MAE =

(x, y)

|X(i, j)

(, )| (13)

Where, M-Width of Image, N-Height of Image
P- Noisy Image , X-Original Image
Table 1 and Table 2 show the comparison of PSNR and MSE for cameraman image at various
noise variancies. Figure 4 and Figure 5 shows that bayes shrinkage has better PSNR and low
MSE than filtering methods and other wavelet based thresholding techniques.
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
40
Table1. Comparison of PSNR for Cameraman image corrupted with Gaussian and Speckle noise
at different Noise variances using db1 (Daubechies Wavelet)


Figure4. Comparison of PSNR for cameraman image (corrupted with Gaussian noise) at
different noise variance

PSNR (PEAK SIGNAL TO NOISE RATIO)
NOISE NOISE
VARIANCE

MEAN
FILTER
MEDIAN
FILTER
UNIVERSAL
THRESHOLD
VISU
SHRINK
BAYES
SHRINK

G
A
U
S
S
I
A
N



N
O
I
S
E



0.001

24.0598

25.4934

27.2016

28.2978

33.7031

0.002

23.2251

24.3480

25.1748

26.1439

29.9001

0.003

22.5261

23.4147

24.0062

24.8430

27.7650

0.004

21.9796

22.6049

23.1590

23.8149

26.0865

0.005

21.4536

22.0205

22.5099

23.0527

25.1235

0.01

19.5569

19.7703

20.3580

20.5660

22.0446










S
P
E
C
K
L
E


N
O
I
S
E



0.001

24.8274

26.6157

28.4073

32.6526

44.0220

0.002

24.5114

26.1260

26.8834

30.4768

40.0535

0.003

24.2207

25.6708

25.9557

29.3585

38.3935

0.004

23.9316

25.2771

25.3274

28.1881

35.6827

0.005

23.7015

24.8599

24.8691

27.5283

34.3460

0.01

22.6357

23.4053

23.3231

25.1853

30.9207
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
41
Table2. Comparison of MSE for Cameraman image corrupted with Gaussian and Speckle noise at
different Noise variances using db1


Figure5. Comparison of MSE for cameraman image (corrupted with Gaussian noise) at different
noise variances



MSE (MEAN SQUARE ERROR)
NOISE NOISE
VARIANCE

MEAN
FILTER
MEDIAN
FILTER
UNIVERSAL
THRESHOLD
VISU
SHRINK
BAYES
SHRINK

G
A
U
S
S
I
A
N



N
O
I
S
E



0.001

255.3265

183.5446

123.8560

96.2288

27.7188

0.002

309.4321

238.9368

197.5136

158.0136

66.5377

0.003

363.4693

296.2178

258.5006

213.1975

108.7875

0.004

412.2133

356.9362

314.1828

270.1428

160.1160

0.005

465.2894

408.3482

364.8271

321.9641

199.8629

0.01

720.1005

685.5656

598.8007

570.7912

406.0842

S
P
E
C
K
L
E


N
O
I
S
E



0.001

213.9645

141.7451

93.8319

35.3036

2.5756

0.002

230.1138

158.6638

133.2721

58.2642

6.4229

0.003

246.0413

176.1971

165.0083

75.3748

9.4130

0.004

262.9796

192.9158

190.6971

98.6903

17.5716

0.005

277.2851

212.3693

211.9193

114.8823

23.9047

0.01

354.4109

296.8613

302.5347

197.0393

52.6035
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
42
The cameraman image is corrupted by gaussian noise of variance 0.01 and results obtained using
filters and wavelets have been shown in Figure 6.


(a) (b) (c)


(d) (e) (f)


(g)
Figure 6. Denoising of cameraman image corrupted by Gaussian noise of variance 0.01
(a) Original image (b) Noisy image (c) Mean Filter (d) Median Filter (e) Universal thresholding
(f) Visu Shrink (g) Bayes shrink

Table 3. shows the comparison of PSNR, MSE and Mean Absolute Error (MAE) for cameraman
image at different decomposition levels. As number of levels increased, PSNR gets decreased
whereas MAE and MSE get increased. Figure 7, 8 and 9 show that decomposition level1 has high
PSNR and low MSE and MAE than other decomposition levels.







The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
43
Table3. Comparison of PSNR, MSE and MAE for Cameraman image corrupted with Gaussian
noise at different decomposition levels using db2

D
E
C
O
M
P
O
S
I
T
I
O
N

L
E
V
E
L




NOISE
VARIANCE

UNIVERSAL
THRESHOLD

VISU SHRINK

PSNR
(db)

MSE

MAE

PSNR
(db)

MSE

MAE

L
E
V
E
L

1



0.001

27.417

117.864

7.9166

28.031

102.305

7.512

0.002

25.483

183.956

10.070

26.028

162.286

9.622

0.003

24.324

240.229

11.632

24.764

217.077

11.235

0.005

22.775

343.185

14.090

23.087

319.419

13.763

0.01

20.610

564.927

18.373

20.740

548.297

18.179

L
E
V
E
L

2


0.001

25.736

173.564

9.612

26.778

136.524

8.767

0.002

23.834

268.933

12.177

24.667

222.007

11.343

0.003

22.673

351.355

14.047

23.477

291.968

13.110

0.005

21.144

499.579

17.007

21.769

432.649

16.135

0.01

19.027

813.403

22.095

19.424

742.403

21.331

L
E
V
E
L

3




0.001

25.201

196.329

10.250

26.473

146.467

9.1091

0.002

23.203

311.007

13.050

24.386

236.814

11.730

0.003

22.037

406.731

15.079

23.109

317.751

13.722

0.005

20.532

575.175

18.176

21.422

468.585

16.764

0.01

18.443

930.519

23.619

19.091

801.536

22.171


The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
44


Figure 7. Comparison of Peak Signal to Noise Ratio (PSNR) for cameraman image (denoising
using Visu Shrink) at different decomposition levels


Figure 8. Comparison of Mean Square Error (MSE) for cameraman image (denoising using Visu
Shrink) at different decomposition levels


Figure 9. Comparison of Mean Absolute Error (MAE) for cameraman image (denoising using
Visu Shrink) at different decomposition levels



The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
45
6. CONCLUSION
In this paper, an analysis of denoising techniques like filters and wavelet methods has been
carried out. Filtering is done by Mean and Median Filter. And three different wavelet
thresholding techniques have been discussed i.e. Universal Thresholding, Bayes Shrink and Visu
Shrink. The results conclude that Bayes shrinkage method has high PSNR at different noise
variance and low MSE. Also the comparison of Wavelet thresholding methods at different
decomposition level has been discussed. From simulation result, it is evident that decomposition
level 1 has high PSNR and low MAE and MSE than other decomposition levels i.e. level 2 and
level 3.This concludes that decomposition level 1 is better in removing Gaussian noise than other
decomposition levels.
REFERENCES
[1] Rajni, Anutam, Image Denoising Techniques An Overview, International Journal of Computer
Applications (0975-8887), Vol. 86, No.16, January 2014.
[2] Akhilesh Bijalwan, Aditya Goyal and Nidhi Sethi, Wavelet Transform Based Image Denoise Using
Threshold Approaches, International Journal of Engineering and Advanced Technology (IJEAT),
ISSN: 2249-8958, Vol.1, Issue 5, June 2012.
[3] S.Arivazhagan, S.Deivalakshmi, K.Kannan, Performance Analysis of Image Denoising System for
different levels of Wavelet decomposition, International Journal of Imaging Science and Engineering
(IJISE), Vol.1, No.3, July 2007.
[4] Pawan Patidar, Manoj Gupta,Sumit Srivastava, Ashok Kumar Nagawat, Image De-noising by
Various Filters for Different Noise, International Journal of Computer Applications, Vol.9, No.4,
November 2010.
[5] Mohammed Ghouse, Dr.M.Siddappa, Adaptive Techniques Based High Impulsive Noise Detection
And Reduction of a Digital Image,Journal of Theoretical and Applied Information Technology.
[6] Jappreet Kaur, Manpreet Kaur, Poonamdeep Kaur, Manpreet Kaur, Comparative Analysis of Image
Denoising Techniques, International Journal of Emerging Technology and Advanced Engineering,
ISSN 2250-2459, Vol. 2, Issue 6, June 2012.
[7] Govindaraj.V, Sengottaiyan.G , Survey of Image Denoising using Different Filters, International
Journal of Science, Engineering and Technology Research (IJSETR) ,Vol.2, Issue 2, February 2013.
[8] Idan Ram, Michael Elad, Generalized Tree-Based Wavelet Transform, IEEE Transactions On
Signal Processing, Vol. 59, No. 9, September 2011.
[9] Rakesh Kumar and B.S.Saini,Improved Image Denoising Techniques Using Neighbouring Wavelet
Coefficients of Optimal Wavelet with Adaptive Thresholding, International Journal of Computer
Theory and Engineering, Vol.4, No.3, June 2012.
[10] Sethunadh R and Tessamma Thomas, Spatially Adaptive image denoising using Undecimated
Directionlet Transform, International Journal of Computer Applications, Vol.84, No. 11,December
2013
[11] S.Kother Mohideen Dr. S. Arumuga Perumal, Dr. M.Mohamed Sathik , Image De-noising using
Discrete Wavelet transform, IJCSNS International Journal of Computer Science and Network
Security, Vol.8 No.1, January 2008.
[12] Savita Gupta, R.C. Chauhan and Lakhwinder Kaur, Image denoising using Wavelet Thresholding,
ICVGIP 2002, Proceedings of the Third Indian Conference on Computer Vision, Graphics Image
Processing, Ahmedabad, India, 2002
[13] S.Grace Chang, Bin Yu, Martin Vetterli , Adaptive Wavelet Thresholding for image denoising and
compression, IEEE Transaction On Image Processing, Vol.9, No.9, September 2000
[14] Nilanjan Dey, Pradipti Nandi, Nilanjana Barman, Debolina Das, Subhabrata Chakraborty , A
Comparative Study between Moravec and Harris Corner Detection of Noisy Images Using Adaptive
Wavelet Thresholding Technique, International Journal of Engineering Research and Applications
(IJERA), ISSN: 2248-9622 , Vol. 2, Issue 1, Jan-Feb 2012.
[15] Tajinder Singh, Rajeev Bedi, A Non - Linear Approach For Image De-Noising Using Different
Wavelet Thresholding, International Journal of Advanced Engineering Research and Studies, ISSN-
2249-8974,Vol.1,Issue3,April-June,2012
The International Journal of Multimedia & Its Applications (IJMA) Vol.6, No.3, June 2014
46
[16] Abdolhossein Fathi and Ahmad Reza Naghsh-Nilchi, Efcient Image Denoising Method Based on a
New Adaptive Wavelet Packet Thresholding Function, IEEE Transaction On Image Processing, Vol.
21, No. 9, September 2012
[17] Virendra Kumar, Dr. Ajay Kumar, Simulative Analysis of Image denoising using Wavelet
ThresholdingTechnique, International Journal of Advanced Research in Computer Engineering and
Technology (IJARCET), Vol.2 , No.5, May 2013
[18] Mark J.T. Smith and Steven L. Eddins, Analysis/SynthesisTechniques for subband image coding,
IEEE Trans. Acoustic Speech and Signal Processing, Vol.38, No.8, Aug 1990
[19] D.L. Donoho and I.M. Johnstone, Denoising by soft thresholding, IEEE Trans. on Information
Theory, Vo.41, 1995
[20] Raghuveer M. Rao, A.S. Bopardikar Wavelet Transforms: Introduction to Theory and Application
published by Addison-Wesley, 2001
[21] S.Sutha, E. Jebamalar Leavline, D. ASR Antony Gnana Singh, A Comprehensive Study on Wavelet
based Shrinkage Methods for Denoising Natural Images, WSEAS Transactions on Signal
Processing, Vol. 9, Issue 4, October 2013
[22] E.Jebamalar Leavline, S.Sutha, D.Asir Antony Gnana Singh, Wavelet Domain Shrinkage Methods
for Noise Removal in Images: A Compendium, International Journal of Computer
Applications,Vol.33, No.10, November 2011
[23] G.Y. Chen, T.D. Bui, A. Krzyak, Image denoising using neighbouring Wavelet coefficients,
Acoustics Speech and Signal processing, IEEE International Conference, Vol.2, May 2004


AUTHORS

Anutam
She is currently pursuing M.Tech from SBS State TechnicalCampus, Ferozepur,
India. She has completed B.Tech from PTU, Jalandhar in 2012. Her areas of interest
includesWireless Communication and Image Processing.



Rajni

She is currently Associate Professor at SBS State Technical Campus Ferozepur,
India. She has completed her M.E. from NITTTR, Chandigarh, India and B.Tech
from REC, NIT, Kurukshetra, India. She has sixteen years of academic experience
and two years industrial experience. She has authored a number of research
papers in International journals, National and International conferences. Her areas
of interest include wireless communication, signal Processing and Antenna design.