International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-4, April 2015
Analysis of Wavelet based Techniques for Image
denoising
Avinash A. Talole, Sachin B. Patel, Dr. Ameya K. Naik,
The methods based on wavelet representations yield very
Abstract— Wavelet based de-noising methods are used for simple algorithms that are often more powerful and easy to
removing the noise present in the image while preserving the work with than traditional methods of function estimation. It
main features of the image. The method is based on threshold consists of decomposing the observed signal into wavelets
estimation for each sub band of the wavelet decomposition of a and using thresholds to select the coefficients, from which a
noise contaminated image, by considering that the sub band
signal is synthesized. There are number of methods are
coefficients have a Speckle noise. The wavelet transformation
(WT) has been used to transform the noisy image, as the main available for threshold detection like VisuShrink,
application of WT is de-noising. We compare methods which NeighShrink, Bayes Shrink, etc. In this paper we deal a new
are based on threshold value. The experiment has been method for Speckle noise methods.
conducted on various test images and compared the The paper is organized as follows: Section II deals with the
performance in terms of peak signal to noise ratio with the general approach to denoising and gives a survey of existing
establish threshold parameters. techniques. Section III discusses the results and analysis of
denoising techniques. This is followed by conclusions.
Index Terms— De-noising, Wavelet transform, VisuShrink,
NeighShrinkSURE, Suggested Method for Speckle noise.
II. DENOISING AND IT’S APPLICATIONS
I. INTRODUCTION
Removing noise from an image effectively and retaining the
One of the main tasks of image processing is to distinguish actual content of the image is still a challenge in field of
between noise and actual contents so that the un-wanted noise Image denoising. Image denoising algorithm consists of few
from the image signal can be removed. The distortion of an steps; consider an input signal and noisy signal .
image by noise is very common that gets introduced during Add these components to get noisy data i.e.
its acquisition, processing, compression, transmission, and
reproduction.
The image usually has noise which is not easily eliminated in
image processing. According to actual image characteristic, Here the noise can be Gaussian, Poisson’s and Salt and
noise statistical property and frequency spectrum distribution pepper, then apply wavelet transform to get .
rule, people have developed many methods of eliminating
noises, which approximately are divided into space and
transformation fields. The space field is data operation →
carried on the original image, and processes the image grey
value, like Neighborhood average method, wiener filter,
center value filter and so on. The transformation field is Modify the wavelet coefficient using different
management in the transformation field of images, and the threshold algorithm and take inverse wavelet transform to get
coefficients after transformation are processed. Then the aim denoising image ̅
of eliminating noise is achieved by inverse transformation,
like wavelet transform [1], [2]. Successful exploitation of →
wavelet transform might lessen the noise effect or even
overcome it completely [3].
In recent years there has been fair amount of work on wavelet The system is expressed in Fig. 1.
thresholding [4-8]. Donoho and Johnstone proposed a
non-linear method to remove the noise [4, 5].This approach is
Input Wavelet Estimate
now widely used in statistics, particularly in signal Noisy Transform Threshold
processing and image analysis. In statistical context this can Image WT
be referred as the estimation of the true curve from data
contaminated with the noise usually assumed to be Gaussian
noise but in this paper we deal with speckle noise.
Denoised Inverse
Image Wavelet
Manuscript received April 22, 2015. Transform
Avinash A. Talole, Department of Electronics & Telecommunication,
K. J. Somaiya College of Engineering, Mumbai, India, 9833474332.
Sachin B. Patel, Department of Electronics & Telecommunication, K. J. Fig. 1: Block diagram of Image denoising using
Somaiya College of Engineering, Mumbai, India, 9892266096. wavelet transform.
Dr. Ameya K. Naik, Department of Electronics & Telecommunication,
K. J. Somaiya College of Engineering, Mumbai, India, 9833275939.
The main task of wavelet thresholding is the selection of
threshold value and the effect of denoising depends on the
417 www.erpublication.org
Analysis of Wavelet based Techniques for Image denoising
selected threshold. A bigger threshold will throw off useful Where T is the threshold, k is the window size and s denotes
information and noise components at the same time while a the sub band.
smaller threshold cannot eliminate the noise effectively. The
C. Suggested Method for Speckle Noise
two thresholding functions frequently used are hard and soft
thresholding. The hard thresholding eliminates coefficients As an alternative to VisuShrink and NeighShrinkSURE, we
that are smaller than a threshold; the soft thresholding shrinks proposed a method [5] for image de-noising using threshold
the coefficients that are larger than the threshold as well. The value
hard and soft thresholding functions are given by (1) and (2).
(10)
̂√ √
| | ̂
{
Where, J is the power of 2 in the sample of
size . This threshold value is assumed based on
{ the experiments on various images of different sizes using the
soft threshold rule. For a given threshold the soft
threshold value is given by,
Where T is the threshold value and w is the wavelet
coefficient. | | (11)
The performance of various speckle reduction techniques is
evaluated using the following standard image quality This is known as shrink or kill rule.
assessment metrics: The proposed threshold value which depends on the size of
the data is asymptotically optical and simpler to implement.
Compared to the universal this threshold value is low hence
Peak Signal to Noise Ratio (PSNR) including some important features in the estimated image.
In all the above estimation process, the noise standard
PSNR= (6) deviation ̂ can be calculated using following equation (12)
Where MSE stands for Mean squared Error between two | |
Images, which is given by ̂ (12)
∑
(7) Where belong to the sub band in the first level of
decomposition, which gives the diagonal details of the image.
Two methods used for denoising are given as follows:
III. RESULTS AND ANALYSIS
A. VisuShrink Following images are the experimental results of the
VisuShrink uses universal threshold given in (8) VisuShrink, NeighShrinkSURE and Suggested method for
Speckle noise. Fig. 2 and Fig. 3 show the output for
VisuShrink for σ=15 and σ=30, Fig. 4 and Fig. 5 show the
√ (8)
output for NeighShrinkSURE for σ=15 and σ=30 whereas
Fig. 6 and Fig. 7 show the output for the suggested method
Where, denotes the noise standard deviation and L is the
for Speckle noise for σ=15 and σ=30.
total number of pixels in an image. This technique yields
overly smoothened images with less preserved details. This is
Original Image Noisy Image
due to the fact that the universal threshold with high
probability yields an estimate that is at least as smooth as the
signal. So the threshold value tends to be high for large values
of L, and that may kill many signal coefficients along with
noise. Hence this does not adapt good with discontinuities in
the signal.
De-Noised Image Using Soft Thresholding
B. NeighShrinkSURE
NeighShrinkSURE, an image denoising method proposed in
[9] is an improved version of NeighShrink. The NeighShrink
uses a suboptimal universal threshold and identical window
size in all wavelet sub-bands, whereas the improved version
of it determines an optimal threshold and neighboring
window size for every sub-band by the Stein’s unbiased risk
estimate (SURE) as given in (8). Fig. 2: Original image, noisy image and De-noised image
using VisuShrink method for σ=15.
( (9)
418 www.erpublication.org
International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-4, April 2015
Original Image Noisy Image Original Image Noisy Image
De-Noised Image Using Soft Thresholding
De-Noised Image Using Soft Thresholding
Fig. 6: Original image, noisy image and De-noised image
Fig. 3: Original image, noisy image and De-noised image
using Suggested method for Speckle noise for σ=15.
using VisuShrink method for σ=30.
Original Image Noisy Image
Original Image Noisy Image
De-Noised Image Using Soft Thresholding
De-Noised Image
Fig. 7: Original image, noisy image and De-noised image
Fig. 4: Original image, noisy image and De-noised image
using Suggested method for Speckle noise for σ=30.
using NeighShrinkSURE method for σ=15.
Original Image Noisy Image VisuShrink, NeighShrinkSURE and Suggested Method for
Spackle noise methods are applied to various test images and
for different wavelets. First we added a Speckle noise to
various test images to obtain the noisy image. While
denoising, we used different wavelets for both methods with
2nd level of decomposition and the results were obtained. In
Table I Results for VisuShrink is given, Table II Results for
NeighShrinkSURE is given and Table III Results for
Suggested Method for Speckle noise is given for different
De-Noised Image wavelets.
Table I: Results for VisuShrink
Wavelet Image/PSNR
Used Barbara Boat Pepper Baboon
db4 21.462 21.964 22.446 20.666
bior2.4 21.513 22.136 22.597 20.751
bior2.6 21.542 22.192 22.626 20.822
Fig. 5: Original image, noisy image and De-noised image
bior6.8 21.541 22.064 22.475 20.658
using NeighShrinkSURE method for σ=30.
419 www.erpublication.org
Analysis of Wavelet based Techniques for Image denoising
First Author – Avinash A. Talole,
Table II: Results for NeighShrinkSURE Post-Graduate Student
Electronics & Telecommunications Engineering,
Wavelet Image/PSNR K. J. Somaiya College of Engineering, Mumbai.
Used Barbara Boat Pepper Baboon
db4 31.788 31.762 32.194 29.653
bior2.4 30.983 31.638 32.193 28.494
bior2.6 31.103 31.72 32.169 28.444
Second Author – Sachin B. Patel,
bior6.8 31.921 31.834 32.321 28.927 Post-Graduate Student
Electronics & Telecommunications Engineering,
K. J. Somaiya College of Engineering, Mumbai.
Table III: Results for Suggested method for Speckle noise
Wavelet Image/PSNR
Used Barbara Boat Pepper Baboon
db4 32.616 32.121 32.328 32.653 Third Author – Dr. Ameya K. Naik,
Asst. Dean(R & D)
bior2.4 32.654 32.155 32.363 32.693
Electronics & Telecommunications Engineering,
bior2.6 32.611 32.115 32.327 32.642 K. J. Somaiya College of Engineering, Mumbai.
bior6.8 32.544 32.044 32.25 32.575
As maximum is the PSNR value more is the co-relation
between original image and de-noised image. From Table
I, Table II and Table III we observe that the PSNR value for
Suggested Method for Speckle noise is higher than
NeighShrinkSURE and VisuShrink.
IV. CONCLUSION
This paper presented a comparison between three Denoising
methods for Speckle noise based on the PSNR value. We
observe that Suggested Method for Speckle noise for
de-noising having a better result than NeighShrinkSURE and
VisuShrink. PSNR value is also higher for Suggested Method
for Speckle noise and the main content of the image are also
retained which usually get affected while denoising.
REFERENCES
[1] Donoho.D.L,Johnstone.I.M, “Ideal spatial adaptation via wavelet
shrinkage”,Biometrika,81,pp.425-455,1994.
[2] GaoZhing, Yu Xiaohai, “Theory and application of MATLAB Wavelet
analysis tools”, National defense industry
publisher,Beijing,pp.108-116,2004.
[3] AglikaGyaourovaUndecimated wavelet transforms for image
denoising,November 19,2002.
[4] D. L. Donoho, “De-Noising by Soft Thresholding,” IEEE Transactions
on Information Theory, Vol. 41, No. 3, 1995, pp.613-627.
[5] Ismail B. and Anjum Khan, “Image De-noising with Stationary
Wavelet Transform,” Vol. 7, Number 2(2013), pp 81-86.
[6] D. L. Donoho and I. M. Johnstone, “Adapting to Un-known
Smoothness via Wavelet Shrinkage,” Journal of American Statistical
Association, Vol. 90, No. 432, 1995, pp. 1200-1224.
[7] M. Jansen, “Noise Reduction by Wavelet Thresholding,” Springer
Verlag Inc., New York, 2001.
[8] Martin Vetterli, S.G, Chang =, BinYu., 2000,”Adaptive wavelet
thresholding for image de-noising and compression”. IEEE
Transaction on image processing 9(9):1532-1546.
[9] Z. Dengwen and C.Wengang, “Denoising with an optimal threshold
and neighboring window”, Pattern Recognition Letters, pp.1694‐
1697,2008.
420 www.erpublication.org
