Вы находитесь на странице: 1из 7

Proceedings oI International ConIerence on Computing Sciences

WILKES100 ICCS 2013


ISBN: 978-93-5107-172-3
Comparative analysis oI SRAD, median Iilter & wavelet transIorm
to de-noise the medical image
Tajinder Kaur
1
, Rajiv Mahajan
2
, Amanpreet Singh
3
, and Anu Bala
4
1
Assistant Professor, Department of Information and Technology, Sant Baba Bhag Singh Institute of Engineering & Technology, Padhiana
2
Professor, Department of Computer Science& Engineering, Global Institute of Emerging & Management Technology, Amritsar
3
Associate Professor, Department of Applied Sciences, Institute of Engineering & Technology, Bhaddal
4
Assistant Professor, Department of Computer Science& Engineering, Sant Baba Bhag Singh Institute of Engineering &Technology, Padhiana
Abstract
Image denoising has remained a primary problem in the field of image processing. This paper provides the comparative analysis
of SRAD, median filter & wavelet, multiscale ridgelet transform. There are many technique used for denoising. Anisotropic,
wavelet & multiscale ridge let is being widely used in reducing the speckle noise of medical image. The performance measure of
image denoising in term of PSNR.
2013 Elsevier Science. All rights reserved.
Keyword: Image Denoising, Multiscale Ridge let, Transform Domain
1. Introduction
Image denoising is a procedure in digital image processing aiming at the removal oI noise. The uncorrupted
image Irom the distorted or noisy image, and is also reIerred to as image 'denoising. There are various methods to
help restore an image Irom noisy distortions. Medical images are oIten aIIected by random noise arising in the
image capturing process. Speckle is a particular kind oI noise which occurs in images obtained by coherent imaging
systems like ultrasound |5|. Image processing is a Iield that continues to grow, with new applications being
developed at an ever increasing pace. It is a Iascinating and exciting area to be involved in today with application
areas ranging Irom the entertainment industry to the space program. One oI the most interesting aspects oI this
inIormation revolution is the ability to send and receive complex data that transcends ordinary written text. Visual
inIormation, transmitted in the Iorm oI digital images, has become a major method oI communication Ior the 21st
century. Ultrasonic images suIIer Irom a special kind oI noise called Speckle |1|. M.Singh et al. described a
comparative study oI various spatial domain Iilters Ior speckle suppression in Ultrasound images |2|.
2. Classification of Denoising Technique
There are two basic methods for image denoising one is Spatial Filtering and another is transform domain method
discuss below.
2.1 Spatial filter:
Spatial filters are direct and high speed processing tools of images. This is the traditional way to remove the noise
from the digital images to employ the spatial filters. We can use spatial filters of different kinds to remove
different kinds of noise. Spatial filter further divided into two parts:
2.1.1. Linear filter
2.1.2 Non linear filter
*
Corresponding author: Er.Tajinder Kaur
226 Elsevier Publications, 2013
Er.Tajinder Kaur, Dr Rajiv Mahajan and Dr. Amanpreet Singh and Er.Anu Bala
2.1.1. Linear filter:
A filter method is linear when the output is a weighted sum of the input pixels. There are many different types of
linear filter but only one type discuss in this paper below:
a) Mean filter is one type of linear filter. A mean filter is the optimal linear filter for Gaussian noise in the sense of
mean square error. Linear filters too tend to blur sharp edges, destroy lines and other fine image details, and perform
poorly in the presence of signal-dependent noise
2.1.2. Non Linear Filter
Non linear spatial filters operate on neighborhoods, and the mechanics of sliding a mask past an image are the
same as was just outlined. Non linear filter further divided below:
a) Median filter: Median filtering is a common step in image processing. Median filter is a well-used nonlinear
filter that replaces the original gray level of a pixel by the median of the gray values of pixels in a specific
neighborhood. The median filter is also called the order specific filter because it is based on statistics derived from
ordering the elements of a set rather than taking the means. This filter is popular for reducing noise without blurring
edges of the image [3].
Advantage of median filter:
Median filter is much less sensitive than mean to extreme value.
Filter is able to remove the outliers, without reducing the sharpness of the image.
b) Speckle Reducing Anisotropic Diffusion (SRAD)
The anisotropic diffusion technique is an extension of conventional Lee filter to suppress the speckle while
preserving the edges Anisotropic is being widely used in reducing speckle noise of medical image. A diffusion
method tailored to ultrasonic and radar imaging application. SRAD is the edge sensitive diffusion for speckled
image. The filter used here is a speckle reduction using anisotropic filter method (SRAD) by Yongiian Yu [4]. In
this sense, the application of this extended version is applied for smoothing the medical ultrasound images in which
signal-dependent, spatially correlated multiplicative noise is present[4].
2.2 Transform Domain:
Transform of signal is just another form of representing the signal. It does not change the information content
present. In image processing there are so many transform but we are discussing only two transform namely Wavelet
transform and multiscale ridge let transform.
2.2.1Wavelet Transform:
Wavelet means a small wave. A wave is an oscillating function of time or space and is periodic. In contrast,
wavelets are localized waves. A wavelet is a wave form of limited duration that has an average value of zero.
Unlike sinusoids that theoretically extend from minus to plus infinity wavelets have a beginning and an end. the
wave and wavelet showing in diagram :
227 Elsevier Publications, 2013

Comparitive Analysis of SRAD,Median Filter and Wavelet transfer to Denoise the Medical Image


(a) (b)
Fig: 1. DiIIerence between wave and wavelet (a) wavelet (b) wave

There are mainly two types of Wavelet Transforms-
1. Continuous Wavelet Transformation (CWT)
2. Discrete Wavelet Transformation (DWT)

Wavelet based Image Denoising: Wavelets are basically mathematical functions which break up the data into
different frequency components, and then we study each component with a resolution matched to its scale [8].
Wavelets are the better technique to handle the different type of noises which is present in an image [10]. They have
advantage over traditional Fourier methods in analyzing physical situations where the signals contain discontinuities
& sharp spikes. The wavelet decomposition of an image is done as follows: In the first level of decomposition, the
image is split into 4subbands, namely the HH, HL, LH and LL sub bands as shown in Figure 2. The HH sub band
gives the diagonal details of the image; the HL sub band gives the horizontal features while the LH sub band
represents the vertical structures[11][12]. The LL sub band is the low resolution residual consisting of low frequency
components and it is this sub band which is further split at higher levels of decomposition [13].

Fig 2: Decomposition oI image


Algorithm:
The basic Procedure for all thresholding method is as follows:
Calculate the DWT of the image.
Threshold the wavelet coefficients (Threshold may be universal or sub band adaptive)
Compute the IDWT to get the denoised estimate.

2.2.2 Multiscale Ridge let Transform:

In transform domain methods, we divide the image in low pass and high pass coefficients. Multiscale ridge lets
based on the ridge let transform combined with a spatial band pass filtering operation to isolate different scales. In
case of multiscale ridge let transform we get the different scale of images and apply A^ trous algorithm to it [6].But
in simple ridge let transform we cannot get the different scale of images.
228 Elsevier Publications, 2013
Er.Tajinder Kaur, Dr Rajiv Mahajan and Dr. Amanpreet Singh and Er.Anu Bala

Algorithm:
Apply the `a trous algorithm with J scales [6].
Apply the radon transform on detail sub-bands of J scales.
Calculate ridge let coefficients by applying 1-D wavelet transform on radon coefficients.
Get the multiscale ridge let coefficients for J scales.

Fig 3.Multiscale ridge let transIorm Ilow graph.|7|

3. Experimental Result

We have implemented and tested filter & transform method on the medical image. Experiments are conducted on
various test images by adding the different types of noise. The PSNR from various methods are compared in table
.Images (Add Gaussian Noise in JPG images)




Table 1: Denoising Results of using filters. In terms of PSNR under Gaussian Noise of Zero Mean and 0.02 Variance.
Image No. SRAD Median
1 21.79 31.41
2 20.61 31.19
3 18.83 32.60
4 25.44 32.90
5 15.98 31.44
Graph 1: Denoising Results of using filters. In terms of PSNR under Gaussian Noise of Zero Mean and 0.02 Variance.



Table 2: Denoising Results of using transform. In terms of PSNR Under Gaussian Noise of Zero Mean and 0.02 Variance.

Image No. Wavelet Multiscale Ridge let
1 23.95 30.53
0
20
40
1 2 3 4 3
lmage no.
S8Au
Medlan
229 Elsevier Publications, 2013
Comparitive Analysis of SRAD,Median Filter and Wavelet transfer to Denoise the Medical Image
2 23.58 29.09
3 16.96 20.48
4 32.57 36.93
5 13.64 13.20
Graph 2: Denoising Results of using transform. In terms of PSNR Under Gaussian Noise of Zero Mean and 0.02 Variance.

Table 3: Denoising Results of using various filters. In terms of PSNR under Speckle Noise of 0.02 Variance.
Image No. SRAD Median
1 34.35 34.27
2 34.40 32.18
3 32.42 33.56
4 31.27 33.45
5 28.23 35.17
Graph 3: Denoising Results of using various filters. In terms of PSNR under Speckle Noise of 0.02 Variance.

Table 4: Denoising Results of using Transform Domain .In terms of PSNR under Speckle Noise of 0.02 variance
Image No. Wavelet Multiscale Ridge let
1 26.82 32.4
2 25.82 29.81
3 18.58 21.6
4 34.25 37.6
5 14.84 16.49
Graph 4: Denoising Results of using Transform Domain .In terms of PSNR under Speckle Noise of 0.02 variance
0
10
20
30
40
1 2 3 4 3
lmage no.
WaveleL
Muluscale
8ldgeleL
0
10
20
30
40
1 2 3 4 3
lmage no.
S8Au
Medlan
230 Elsevier Publications, 2013
Er.Tajinder Kaur, Dr Rajiv Mahajan and Dr. Amanpreet Singh and Er.Anu Bala



4. Conclusions
To Denoising the medical image with SRAD, Median Iilter and Wavelet, multiscale ridge let transIorm. The
results oI median Iilter are best to remove the Gaussian noise. But when it is required to remove speckle noise,
SRAD Iilter is considered. By applying the wavelet and multiscale ridge let transIorm on noisy image .the outcome
oI multiscale ridge let is best as compared to wavelet transIorm.

References

|1| Somatic Udomhunsakul` and Pichet Wongsita 'Ultrasonic speckle denoising using the combination oI wavelet transIorm and wiener
Iilter,IEEE(2004)
|2| M. Singh, Dr.S.Singh,Dr. S.Kansal, 'Comparative Analysis oI spatial Iilters Ior speckle reduction in ultrasound images, World Congress on
Computer Science and InIormation Engineering IEEE (2008).
|3| Mariana Carmen Nicolae, Lumini|a Moraru, Laura Onose, Comparative Approach For Speckle Reduction In Medical Ultrasound Images.
|4| Yongjian Yu and Scott T. Acton, "Speckle Reducing Anisotropic DiIIusion", IEEE Transactions on Image Processing, page(s): 1260-1270,
No. 11, Vol.11, Nov 2002
|5| Arash Vosoughi, Mohammad. B. Shamsollahi 'Speckle Noise Reduction oI Ultrasound Images Using M-band Wavelet TransIorm and
Wiener Filter in a Homomorphic Framework International ConIerence on BioMedical Engineering and InIormatics IEEE 2008.
|6| Yong-bing Xu, Chang-Sheng Xie, and Cheng-Yong Zheng, 'An Application oI the a Trous Algorithm in Detecting InIrared Targets, IEEE
conI. on Wavelet Analysis and Pattern Recognition, Beijing, China, 2-4: Nov. 2007, pp.1015-1019
|7| Jean-Luc starck, Emmanual J.Candes, and David L.Donoho,The Curvelet transIorm Ior image Denoising IEEE Trans.Image
processing,vol.11,no.6,pp.670-684,2002
|8| A.Kaur and K.Singh 'Speckle Noise Reduction by using Wavelets 'NCCI 2010 - INDIA, 19-20 March 2010.
|9| 'IEEE Computational Science and Engineering, summer , vol. 2, num.2, Published by the IEEE Computer Society, 10662 Los Vaqueros
Circle, Los Alamitos, CA 90720, USA.
|10| D.L.Donoho and I. M. Johnstone, "Denoising by soIt thresholding, IEEE Trans. on IriIorm. Theory, Vol. 41, pp. 613-627, 1995 |12|
S.Sudha, G.R.Suresh and R.Sukanesh, 'wavelet based image denoising using Adaptive thresholding the International ConIerence on
Computational Intelligence and Multimedia Applications IEEE (2007).
0
20
40
1 2 3 4 3
lmage no.
WaveleL
231 Elsevier Publications, 2013
Index

I
Image denoising, 226, 228

M
Multiscale ridge let, 228229

T
Transform domain, 226228

Вам также может понравиться