Hindawi Publishing Corporation

Mathematical Problems in Engineering

Volume 2015, Article ID 280251, 9 pages
http://dx.doi.org/10.1155/2015/280251

Research Article
A New Wavelet Threshold Determination Method Considering
Interscale Correlation in Signal Denoising

Can He,1 Jianchun Xing,1 Juelong Li,2 Qiliang Yang,1 and Ronghao Wang1
1
College of Defense Engineering, PLA University of Science and Technology, Nanjing 210007, China
2
Technical Management Office of Naval Defense Engineering, Beijing 100841, China

Correspondence should be addressed to Jianchun Xing; [email protected]

Received 28 August 2014; Revised 8 October 2014; Accepted 8 October 2014

Academic Editor: Zheng-Guang Wu

Copyright © 2015 Can He et al. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Due to simple calculation and good denoising effect, wavelet threshold denoising method has been widely used in signal denoising.
In this method, the threshold is an important parameter that affects the denoising effect. In order to improve the denoising effect of
the existing methods, a new threshold considering interscale correlation is presented. Firstly, a new correlation index is proposed
based on the propagation characteristics of the wavelet coefficients. Then, a threshold determination strategy is obtained using the
new index. At the end of the paper, a simulation experiment is given to verify the effectiveness of the proposed method. In the
experiment, four benchmark signals are used as test signals. Simulation results show that the proposed method can achieve a good
denoising effect under various signal types, noise intensities, and thresholding functions.

1. Introduction coefficients and then using the processed coefficients to

reconstruct signals. Among these methods, wavelet threshold
Due to the combined impacts of the internal measurement method has been used the most widely because of its simple
system and the external environmental factors, measured calculation and good effect.
signals are often contaminated by noise [1]. Therefore, signal Wavelet threshold method was proposed by Donoho
denoising technology has been a hot topic in the field of and Johnstone [8], whose main idea is to reconstruct signal
signal processing. In a noisy signal, noise energy is generally on the basis of thresholding coefficients. The denoising
concentrated in the high frequency region, and the spectrum effect of wavelet threshold method depends on threshold
of useful signal is distributed in the low frequency region determination. If the selected threshold is too large, then
[2]. According to this theory, a variety of signal denoising some useful information is filtered out, and if the threshold
methods were put forward, such as mean filtering [3], is too small, then a certain amount of noise is retained. In
median filtering [4] and Wiener filtering [5]. Most of these order to solve this problem, many researchers studied the
methods can be considered as low-pass filters essentially. threshold determination methods. Donoho and Johnstone
However, in some signals, the high frequency region not only [8] presented a universal threshold by analyzing normal
contains noise but also possesses a lot of useful information. Gaussian noise model; Tao et al. [9] improved the univer-
Therefore, directly filtering out high frequency information is sal threshold and indicated that the threshold should be
unreasonable. Recently, because of multiresolution and low adaptively changed with the scale alter. The defect in this
entropy, wavelet transform has become a popular research kind of methods is that universal threshold is often set too
topic in the signal denoising field. A number of methods large, which may lead to overkill the useful information.
based on wavelet theory have been proposed, such as wavelet Chang et al. [10] assumed that the wavelet coefficients
coefficient modulus maxima method [6], wavelet correlation obeyed generalized Gaussian distribution and proposed a
method [7], and wavelet threshold method [8]. The essence Bayesian threshold method; Lu and Loizou [11] considered
of these methods is nonlinear processing on the wavelet the coefficients obeyed Gaussian distribution and presented
2 Mathematical Problems in Engineering

a new threshold based on maximum a posteriori probability; 𝑋(𝑘) still contain two parts. One part is from the original
Li et al. [12] supposed the coefficients obeyed generalized signal 𝑆(𝑘), and the other part is brought from the noise 𝐸(𝑘).
Gamma distribution and put forward a threshold method Wavelet transform can concentrate signal energy on some
based on Bayesian shrinkage. All of these methods are based large wavelet coefficients and distribute the noise energy
on a particular coefficient distribution, but it may not satisfy throughout the whole wavelet domain. Thus, large amplitude
the distribution to a specific signal. Donoho and Johnstone wavelet coefficients may be produced by the useful signal, and
[13] proposed a new threshold method based on minimax the small amplitude is likely to represent the noise. According
criterion. However, this method requires the prior knowledge to this characteristics of wavelet coefficients, Donoho and
of the original signal, while the information of original signal Johnstone [8] proposed wavelet threshold method, which can
is difficult to obtain in reality. Stein’s unbiased risk estimate be divided into three steps:
(SURE) criterion [8] and generalized cross validation (GCV) (1) choosing the appropriate wavelets basis and decom-
criterion [14] were presented based on the idea of parameter position scale and computing the corresponding
estimates, in which SURE criterion is unbiased estimate wavelet coefficients,
of the minimized mean square error (MSE) criterion, and
GCV criterion is biased estimates of the minimized MSE (2) selecting the proper threshold and thresholding func-
criterion. Cai and Zhou [15] proposed a data-driven threshold tion and obtaining the estimated values of the wavelet
determination method based on SURE criterion. Autin et al. coefficients,
[16] put forward a new idea by combining different threshold (3) reconstructing the signal based on the estimated
rules. values of wavelet coefficients by inverse wavelet trans-
Although the wavelet threshold method has developed form.
significantly, some deficiencies remain. The primary defect is In the wavelet threshold denoising method, a core issue
that most of these methods ignore the relationship between is to determine the optimal threshold. Threshold can make a
the wavelet coefficients. According to wavelet correlation great influence on the denoising effect. If the threshold value
theory [7], the wavelet coefficients of a useful signal have a is too small, then considerable noise will still exist, and if the
strong correlation in various decomposition scales, whereas threshold value is too large, then some important feature of
the wavelet coefficients of noise are weakly correlated or signal may be filtered out. As mentioned in the introduction,
uncorrelated. Therefore, analysis of the coefficients corre- many existing methods can determine the threshold. Among
lation can help distinguish useful information or noise. these methods, universal threshold is the most widely used
However, the existing threshold methods are mostly from the because of its simpleness and effectiveness. The formula for
minimum error criterion or other optimization criteria that the universal threshold is expressed as follows:
do not consider the coefficient correlation, and this defect
may diminish the signal denoising effect. 𝜆 = 𝜎√2 ln (𝑁), (2)
The goal of this paper is to propose a new wavelet where 𝜎 is the average variance of the noise and 𝑁 is the signal
threshold determination method that considers the interscale length.
correlation of coefficients. The method firstly adopts the 𝜎 is calculated using median estimate method. The
universal threshold as the basic threshold. And then, a formula is as follows:
new correlation index is presented based on the wavelet 󵄨 󵄨
Median (󵄨󵄨󵄨𝑊1,𝐾 󵄨󵄨󵄨)
correlation theory. At last, the new threshold is obtained 𝜎= , (3)
using the correlation index to improve the basic threshold. 0.6745
The remaining parts of the paper are organized as follows. where 𝑊1,𝑘 represent all the wavelet coefficients in scale 1.
Section 2 introduces the wavelet threshold denoising theory. Because universal threshold is the most widely used
Section 3 proposes a new correlation index and presents a threshold method, therefore it is selected as basis to construct
new threshold determination method based on the new index the new threshold in this paper.
and basic threshold. Section 4 provides a simulated experi- Thresholding function reflects different estimation strate-
ment using four benchmark signals to verify the effectiveness gies to the wavelet coefficients. There are two well-known
of the new threshold. The conclusion is given in Section 5. thresholding functions named hard thresholding function
and soft thresholding function. Their main ideas are to both
2. Wavelet Threshold Denoising Theory remove small wavelet coefficients and shrink large wavelet
coefficients. Hard thresholding function shown in Figure 1(a)
In one-dimension noisy signals, noise acts on the original is defined as
󵄨󵄨 󵄨󵄨
{𝑊𝑗,𝑘 ; 󵄨󵄨󵄨𝑊𝑗,𝑘 󵄨󵄨󵄨 ≥ 𝜆
signal through linear superposition:
𝑊̂ 𝑗,𝑘 = { (4)
𝑋 (𝑘) = 𝑆 (𝑘) + 𝐸 (𝑘) , (1) 󵄨󵄨 󵄨󵄨
{0; 󵄨
󵄨 𝑊 󵄨
󵄨 < 𝜆.
󵄨 𝑗,𝑘 󵄨
where 𝑋(𝑘) is the noisy signal, 𝑆(𝑘) is the original signal, and Soft thresholding function shown in Figure 1(b) is defined
𝐸(𝑘) is white Gaussian noise, subject to 𝑁 (0, 𝜎2 ) distribu- as
󵄨 󵄨 󵄨 󵄨
tion.
̂ sgn (𝑊𝑗,𝑘 ) (󵄨󵄨󵄨󵄨𝑊𝑗,𝑘 󵄨󵄨󵄨󵄨 − 𝜆) ; 󵄨󵄨󵄨󵄨𝑊𝑗,𝑘 󵄨󵄨󵄨󵄨 ≥ 𝜆
Wavelet transform is a linear transform. Therefore, 𝑊𝑗,𝑘 = { 󵄨󵄨 󵄨 (5)
0; 󵄨󵄨𝑊𝑗,𝑘 󵄨󵄨󵄨 < 𝜆.
wavelet coefficients obtained through wavelet transform of 󵄨 󵄨
Mathematical Problems in Engineering 3

̂
W ̂
W

−𝜆 𝜆 W −𝜆 𝜆 W

(a) Hard thresholding function (b) Soft thresholding function

Figure 1: Hard and Soft thresholding functions.

Some researchers focused on studying the thresholding 3.2. A New Interscale Correlation Index. Crouse et al. [19]
function and proposed some new functions [17, 18]. Because found that the wavelet coefficients typically had the following
the core issue of this study is to determine the threshold, the distribution characteristics:
classical hard and soft thresholding functions are selected in
(1) clustering: if a particular wavelet coefficient is
the experimental section.
large/small, then adjacent coefficients are very likely
to also be large/small;
3. A New Wavelet Threshold (2) persistence across scale: large/small values of wavelet
Determination Method coefficients tend to propagate across scales.

3.1. Wavelet Correlation Denoising Theory. The propagation According to this theory, a new index is proposed to
characteristics of wavelet coefficients show that signal has measure the interscale correlation of wavelet coefficients in
a strong correlation in various decomposition scales, while this paper:
noise is weakly correlated or uncorrelated. Considering this max |𝑊 (:, 𝑛)| − min |𝑊 (:, 𝑛)|
feature, Xu et al. [7] proposed a classical wavelet correlation 𝐾 (𝑛) = , (7)
min |𝑊 (:, 𝑛)|
denoising method named spatially selective noise filtration.
In this method, the correlation factors of different scales are where 𝑊(:, 𝑛) represents all the wavelet coefficients in loca-
calculated firstly. And then, the larger values are retained as tion point 𝑛.
useful information by comparing the correlation factors with When 𝐾(𝑛) ∈ [0, 𝑟), the difference between maximum
original coefficients. Finally, the signal is reconstructed using and minimum of the wavelet coefficients in location point
the inverse wavelet transform. 𝑛 is little. It is considered that there is a strong correlation
In this method, the correlation of wavelet coefficients is of wavelet coefficients in location point 𝑛. Therefore, the
measured using the normalised correlation index 𝑁Cor: location point 𝑛 is likely to be a signal point. A smaller value
of the 𝐾(𝑛) means a bigger possibility to be a signal point
for location point 𝑛. When 𝐾(𝑛) ∈ [𝑟, +∞), the difference
𝑃𝑊 (𝑗) between maximum and minimum of location point 𝑛 is big.
𝑁Cor (𝑗, 𝑛) = Cor (𝑗, 𝑛) √ , (6)
𝑃Cor (𝑗) A weak correlation of wavelet coefficients is considered in
location point 𝑛. Therefore, the location point 𝑛 is likely to
be a noise point.
where Cor(𝑗, 𝑛) = 𝑊𝑗,𝑛 𝑊𝑗+1,𝑛 represents the coefficients 𝑟 is an important parameter to measure the coefficients
correlation between 𝑊𝑗,𝑛 and 𝑊𝑗+1,𝑛 ; 𝑊𝑗,𝑛 represents the correlation. If the value is selected too small, it may result in
coefficient in location point 𝑛 and scale 𝑗; 𝑃Cor(𝑗) = overkilling some useful signal points. If the value is chosen
∑𝑁 2
𝑛=1 Cor(𝑗, 𝑛) represents the correlation factor energy of too large, it may lead to retain some noise points. As a lot
scale 𝑗; 𝑃𝑊(𝑗) = ∑𝑁 2
𝑛=1 𝑊𝑗,𝑛 represents the coefficients energy of experiments shown, when 𝑟 takes 0.5 to 1.5, the denoising
of scale 𝑗. effect is good. Therefore, in the experiment of this paper, 𝑟 is
𝑁Cor can describe the correlation among the wavelet set 1.
coefficients to a certain extent, but its calculation is compli-
cated and inefficient. Therefore, it is necessary to propose a 3.3. A New Threshold Determination Strategy Considering
simple and efficient correlation index. Interscale Correlation. 𝐾 is an index measuring the interscale
4 Mathematical Problems in Engineering

correlation of wavelet coefficients, which can represent the NS is the noisy signal, DSUT is reconstructed signal using
possibility of a certain point belonging to the signal point. universal threshold, DSPT expresses the reconstructed signal
Because universal threshold often has the risk of overkilling using the proposed threshold.
the useful information, therefore, when 𝐾(𝑛) ∈ [0, 𝑟), The denoising results of soft thresholding function are
the threshold needs to be shrunk to retain location point given in Figures 4 and 5.
𝑛, and when 𝐾(𝑛) ∈ [𝑟, +∞), the threshold of location In order to exclude the effect of noise intensity, experi-
point 𝑛 remains invariant. Based on a lot of experiments, a ment is retested with 𝜎 = 3 white Gaussian noise.
new threshold determination method considering interscale
correlation is presented as follows: 4.2. Result Analysis. As shown in Figure 2, when 𝜎 = 5
and hard thresholding function is selected, denoising signal
{ 0.7𝜆 𝑘 (𝑛) ∈ [0, 0.5𝑟)
{
{ obtained by the proposed method is closer to the original
{0.8𝜆 𝑘 (𝑛) ∈ [0.5𝑟, 0.8𝑟)
𝑇new ={ (8) signal than universal threshold for all the four different
{
{ 0.9𝜆 𝑘 (𝑛) ∈ [0.8𝑟, 𝑟) signals. As shown in Figure 3, the proposed method can
{
{𝜆 𝑘 (𝑛) ∈ [𝑟, +∞) , achieve a smaller MSE and a larger SNR than traditional
method. In order to avoid the interference of threshold-
where 𝜆 is universal threshold. ing function, Figures 4 and 5 show the denoising results
If 𝐾(𝑛) ∈ [0, 0.5𝑟), then the correlation of the wavelet under soft thresholding function. Comparing Figure 2 with
coefficients is considered to be very large, and the location Figure 4, denoising signal obtained by our method is closer
𝑛 is very likely to be a signal point. Therefore, we set the to the original signal than universal threshold, regardless of
threshold value 𝑇new = 0.7𝜆. If 𝐾(𝑛) ∈ [0.5𝑟, 0.8𝑟), then the which hard threshold function or soft threshold function
correlation of the wavelet coefficients is considered large, and is selected. From Figures 3 and 5, we find that no matter
the location point 𝑛 is likely to be a signal point. Therefore, which thresholding function is selected, denoising signal
we set the threshold value 𝑇new = 0.8𝜆. If 𝐾(𝑛) ∈ [0.8𝑟, 𝑟), using our method can obtain a smaller MSE and greater SNR.
then the correlation of the wavelet coefficients is considered Experiment is retested with 𝜎 = 3 white Gaussian noise to
to be slightly large, and there is slight possibility for location exclude the effect of noise intensity. The experiment results
𝑛 to be a signal point. Therefore, we set the threshold value are shown in Figures 6, 7, 8, and 9, which indicate that
𝑇new = 0.9𝜆. If 𝐾(𝑛) ∈ [𝑟, +∞), the correlation of the wavelet the denoising effect and index of the proposed method are
coefficients is considered very little, and there is almost no superior to those of the universal threshold with both heavy
possibility for location 𝑛 to be a signal point. Therefore, we and light noise.
set the threshold value 𝑇new = 𝜆.
5. Conclusion
4. Simulation Experiment
A new method considering interscale correlation is presented
4.1. Experimental Parameter Settings and Experiment Results. to solve the problem of wavelet threshold determination.
In order to verify the validity of the new threshold, four classic Firstly, a new index is proposed to measure the coefficients
benchmark signals, namely, Blocks, Bumps, Heavy Sine, and correlation. Then, a new threshold determination strategy is
Doppler, are used as test signals. The length of the signal is 512 obtained using new index to improve universal threshold.
and 𝑟 is set 1. Two strengths of white Gaussian noise 𝜎 = 3 and Some conclusions are summarized as follows.
𝜎 = 5 are added to the test signals in the experiments. The
calculation process of the correlation index 𝐾 requires that (1) According to the propagation characteristics of the
the number of wavelet coefficients to be the same in different wavelet coefficients, a new index is proposed to
decompositions. Therefore, stationary wavelet transform is measure the coefficients correlation. Compared with
used in this paper [20]. Wavelet basis adopts sym6 wavelet, traditional index 𝑁Cor, the new index has the advan-
wavelet decomposition scale is set to be 3, and hard and soft tage of a simple structure and convenient calculation.
thresholding functions are chosen as thresholding functions.
The traditional universal threshold is selected as comparative (2) Universal threshold has the defect of overkilling the
experiment. In order to measure the denoising effect under useful information. In order to address the issue,
different thresholds, MSE and signal-to-noise-ratio (SNR) are interscale correlation is used to shrink the univer-
selected as comparative index: sal threshold. Experimental results show that the
proposed method can achieve optimum denoising
∑𝑁
𝑘=1 [𝑋 (𝑘) − 𝑆 (𝑘)]
2
effect under different signal types, noise intensities,
MSE = , and thresholding functions. Therefore, this method is
𝑁
(9) effective and superior in signal denoising.
∑𝑁
𝑘=1 𝑆(𝑘)
2
SNR = 10 ln .
∑𝑁
𝑘=1 [𝑋 (𝑘) − 𝑆 (𝑘)]
2 Universal threshold is selected as the basic threshold in
this paper. In theory, the proposed idea is also applicable
Figures 2 and 3 show the denoising effects of the four test to the other threshold methods. Future research will be
signals with 𝜎 = 5 Gaussian white noise and hard threshold- conducted on using the proposed idea to improve other
ing function. In the figures, OS represents the original signal, thresholds.
Mathematical Problems in Engineering 5

14 20
12
10 15

Amplitude
Amplitude

10
6
4
5
2
0 0
−2
−4 −5
0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples

NS DSUT NS DSUT
OS DSPT OS DSPT
(a) Block signal (b) Bumps signal
10 10

5 5

0
Amplitude

0
Amplitude

−5 −5

−10 −10

−15 −15
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples

NS DSUT NS DSUT
OS DSPT OS DSPT
(c) Heavy sine signal (d) Doppler signal

Figure 2: Denoising effect of 𝜎 = 5, hard thresholding function.

0.5 23
0.45 22
0.4
21
0.35
0.3 20
MSE

SNR

0.25 19
0.2 18
0.15
17
0.1
0.05 16
0 15
Blocks Bumps Heavy sine Doppler Blocks Bumps Heavy sine Doppler
𝜎 = 5, hard thresholding function 𝜎 = 5, hard thresholding function

Universal thresholding Universal thresholding

Proposed thresholding Proposed thresholding
(a) MSE (b) SNR

Figure 3: Denoising index of 𝜎 = 5, hard thresholding function.

6 Mathematical Problems in Engineering

14 20
12
10 15

8
10

Amplitude
Amplitude

6
4
5
2
0 0
−2
−4 −5
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples

NS DSUT NS DSUT
OS DSPT OS DSPT
(a) Block signal (b) Bumps signal
10 10

5 5

0 0
Amplitude
Amplitude

−5 −5

−10 −10

−15 −15
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples
NS DSUT NS DSUT
OS DSPT OS DSPT
(c) Heavy sine signal (d) Doppler signal

Figure 4: Denoising effect of 𝜎 = 5, soft thresholding function.

0.7
22
0.6

0.5 20

0.4 18
MSE

SNR

0.3
14
0.2
12
0.1

0 10
Blocks Bumps Heavy sine Doppler Blocks Bumps Heavy sine Doppler
𝜎 = 5, soft thresholding function 𝜎 = 5, soft thresholding function

Universal thresholding Universal thresholding

Proposed thresholding Proposed thresholding
(a) MSE (b) SNR

Figure 5: Denoising index of 𝜎 = 5, soft thresholding function.

Mathematical Problems in Engineering 7

14 12
12 10
10
8
8
6
Amplitude

Amplitude
4 4
2 2
0
0
−2
−4 −2
−6 −4
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples
NS DSUT NS DSUT
OS DSPT OS DSPT
(a) Block signal (b) Bumps signal
8 8
6 6
4 4
2
2
Amplitude

0
Amplitude

0
−2
−2
−4
−6 −4

−8 −6
−10 −8
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples
NS DSUT NS DSUT
OS DSPT OS DSPT
(c) Heavy sine signal (d) Doppler signal

Figure 6: Denoising effect of 𝜎 = 3, hard thresholding function.

0.35 20
19.5
0.3
19
0.25 18.5

0.2 18
MSE

SNR

17.5
0.15 17
0.1 16.5
16
0.05
15.5
0 15
Blocks Bumps Heavy sine Doppler Blocks Bumps Heavy sine Doppler
𝜎 = 3, hard thresholding function 𝜎 = 3, hard thresholding function

Universal thresholding Universal thresholding

Proposed thresholding Proposed thresholding
(a) MSE (b) SNR

Figure 7: Denoising index of 𝜎 = 3, hard thresholding function.

8 Mathematical Problems in Engineering

14 12
12 10
10
8
8
6
Amplitude

Amplitude
4 4
2 2
0
0
−2
−4 −2
−6 −4
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples

NS DSUT NS DSUT
OS DSPT OS DSPT
(a) Block signal (b) Bumps signal
8 8
6 6
4 4
2
2
Amplitude

Amplitude

0
0
−2
−2
−4
−6 −4

−8 −6
−10 −8
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Samples Samples

NS DSUT NS DSUT
OS DSPT OS DSPT
(c) Heavy sine signal (d) Doppler signal

Figure 8: Denoising effect of 𝜎 = 3, soft thresholding function.

0.7 20
19
0.6
18
0.5 17

0.4 16
SNR
MSE

15
0.3 14
0.2 13
12
0.1
11
0 10
Blocks Bumps Heavy sine Doppler Blocks Bumps Heavy sine Doppler
𝜎 = 3, soft thresholding function 𝜎 = 3, soft thresholding function

Universal thresholding Universal thresholding

Proposed thresholding Proposed thresholding
(a) MSE (b) SNR

Figure 9: Denoising index of 𝜎 = 3, soft thresholding function

Mathematical Problems in Engineering 9

Conflict of Interests [15] T. T. Cai and H. H. Zhou, “A data-driven block thresholding

approach to wavelet estimation,” The Annals of Statistics, vol. 37,
The authors declare that there is no conflict of interests no. 2, pp. 569–595, 2009.
regarding the publication of this paper. [16] F. Autin, J.-M. Freyermuth, and R. von Sachs, “Combining
thresholding rules: a new way to improve the performance of
wavelet estimators,” Journal of Nonparametric Statistics, vol. 24,
References no. 4, pp. 905–922, 2012.
[1] T.-H. Yi, H.-N. Li, and X.-Y. Zhao, “Noise smoothing for [17] A. Fathi and A. R. Naghsh-Nilchi, “Efficient image denoising
structural vibration test signals using an improved wavelet method based on a new adaptive wavelet packet thresholding
thresholding technique,” Sensors, vol. 12, no. 8, pp. 11205–11220, function,” IEEE Transactions on Image Processing, vol. 21, no. 9,
2012. pp. 3981–3990, 2012.
[2] A. G. Bruce, D. L. Donoho, H. Y. Gao, and R. D. Martin, [18] R. Yang and M. Ren, “Wavelet denoising using principal
“Denoising and robust nonlinear wavelet analysis,” in Optical component analysis,” Expert Systems with Applications, vol. 38,
Engineering and Photonics in Aerospace Sensing, Proceedings of no. 1, pp. 1073–1076, 2011.
SPIE, pp. 325–336, Orlando, Fla, USA, 1994. [19] M. S. Crouse, R. D. Nowak, and R. G. Baraniuk, “Wavelet-based
[3] X. Zhang and Y. Xiong, “Impulse noise removal using direc- statistical signal processing using hidden Markov models,” IEEE
tional difference based noise detector and adaptive weighted Transactions on Signal Processing, vol. 46, no. 4, pp. 886–902,
mean filter,” IEEE Signal Processing Letters, vol. 16, no. 4, pp. 1998.
295–298, 2009. [20] S. Zhong and S. O. Oyadiji, “Crack detection in simply sup-
[4] G. Gupta, “Algorithm for image processing using improved ported beams using stationary wavelet transform of modal
median filter and comparison of mean, median and improved data,” Structural Control and Health Monitoring, vol. 18, no. 2,
median filter,” International Journal of Soft Computing and pp. 169–190, 2011.
Engineering, vol. 1, no. 5, pp. 2231–2307, 2011.
[5] K.-M. Chang and S.-H. Liu, “Gaussian noise filtering from ECG
by Wiener filter and ensemble empirical mode decomposition,”
Journal of Signal Processing Systems, vol. 64, no. 2, pp. 249–264,
2011.
[6] B. Le, Z. Liu, and T. Gu, “Weak LFM signal detection based
on wavelet transform modulus maxima denoising and other
techniques,” International Journal of Wavelets, Multiresolution
and Information Processing, vol. 8, no. 2, pp. 313–326, 2010.
[7] Y. Xu, J. B. Weaver, D. M. Healy Jr., and J. Lu, “Wavelet transform
domain filters: a spatially selective noise filtration technique,”
IEEE Transactions on Image Processing, vol. 3, no. 6, pp. 747–758,
1994.
[8] D. L. Donoho and I. M. Johnstone, “Ideal spatial adaptation by
wavelet shrinkage,” Biometrika, vol. 81, no. 3, pp. 425–455, 1994.
[9] Z. Tao, H.-M. Zhao, X.-J. Zhang, and D. Wu, “Speech enhance-
ment based on the multi-scales and multi-thresholds of the
auditory perception wavelet transform,” Archives of Acoustics,
vol. 36, no. 3, pp. 519–532, 2011.
[10] S. G. Chang, B. Yu, and M. Vetterli, “Adaptive wavelet threshold-
ing for image denoising and compression,” IEEE Transactions on
Image Processing, vol. 9, no. 9, pp. 1532–1546, 2000.
[11] Y. Lu and P. C. Loizou, “Estimators of the magnitude-squared
spectrum and methods for incorporating SNR uncertainty,”
IEEE Transactions on Audio, Speech, and Language Processing,
vol. 19, no. 5, pp. 1123–1137, 2011.
[12] H.-C. Li, W. Hong, Y.-R. Wu, and P.-Z. Fan, “Bayesian wavelet
shrinkage with heterogeneity-adaptive threshold for SAR image
despeckling based on generalized gamma distribution,” IEEE
Transactions on Geoscience and Remote Sensing, vol. 51, no. 4,
pp. 2388–2402, 2013.
[13] D. L. Donoho and I. M. Johnstone, “Minimax risk over 𝑙𝑝 -balls
for 𝑙𝑝 -error,” Probability Theory and Related Fields, vol. 99, no. 2,
pp. 277–303, 1994.
[14] M. Jansen and A. Bultheel, “Multiple wavelet threshold estima-
tion by generalized cross validation for images with correlated
noise,” IEEE Transactions on Image Processing, vol. 8, no. 7, pp.
947–953, 1999.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

Function Spaces
Hindawi Publishing Corporation
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014