Comparison of Wavelet Watermarking Method With & without Estimator Approach
1 like465 views
1. The document compares a wavelet watermarking method with and without an estimator approach for improving robustness against noise attacks. 2. Using an M-estimator at extraction improves imperceptibility and robustness by estimating and rejecting outlier pixels caused by noise. 3. Statistical analysis on watermarked images subjected to noise attacks shows the estimator approach reduces MSE and increases PSNR and correlation, indicating superior extraction quality compared to the standard wavelet method without estimator.
![IJSRD - International Journal for Scientific Research & Development| Vol. 1, Issue 3, 2013 | ISSN (online): 2321-0613
All rights reserved by www.ijsrd.com 468
Comparison of Wavelet Watermarking Method With & without
Estimator Approach
Neha Y. Joshi1
Kavindra R. Jain2
Pratyaksh A Maru3
1
Research Scholar
2
Assistant Professor, G. H. Patel College of Engg. and Tech., Vallabh Vidyanagar, India
3
Assistant Professor, Dr. Jivraj Mehta Institute of Tech, Mogar, India
Abstract — In this paper we propose an Estimator approach
with wavelet watermarking method which is capable to hide
watermark in the host image based on wavelet domain
technique. Using the Estimator approach the proposed
technique becomes robust against different noise attacks.
For the evaluation of Imperceptibility & Robustness of the
proposed method we have calculated basic statistical
parameters. We have tested watermarked image against
different noise attacks at different noise densities. Due to the
use of estimator the perceptible quality of extracted image is
better though the image is degraded by high density noise.
I. INTRODUCTION
Nowadays technologies are changing so to use digital
watermarking instead of the age old techniques for hiding
information & protecting multimedia data has become the
greatest area of interest. This makes the availability of
digital information in the form of audio, video & images are
quiet easily and immensely available and in reach to public
via web. The main concern in such techniques is the
robustness & imperceptibility.
The watermarking can be applied to the multimedia
data like Images, text, audio & video. Watermarking
technique mainly contains two processes embedding &
extracting for that different types of watermark & various
watermarking techniques are available. According to the
properties of watermark it can be divided into two main
categories visible & invisible watermark[2]. Human audio &
video system means Human visual perception decides the
properties of watermark. Watermarking technique can also
be classified based on the extraction process. According to
it, we can divide it into three types Blind, Semi-Blind &
Non-Blind. Techniques which do not require the original
data or signal fall into the first category which is blind.
Techniques which require the original watermark are
considered as a Semi-Blind watermarking & techniques
which require the original signal for extraction process is the
Non-Blind technique. According to the domain we can also
classified watermarking as a spatial & Frequency domain. In
frequency domain we cannot directly embed watermark for
that first we have to convert original signal & watermark
into frequency domain using different transforms while in
spatial domain we can directly apply watermark by
changing pixel values or by using spread spectrum approach
[3]. The past researches suggest that compare to spatial
domain; frequency domain watermarking is more robust
against different attacks. To achieve data protection the
watermarking technique should be imperceptible with high
level of security i.e. watermarked image should not reveal
any things about hiding information. The whole paper has
been divided into V sections. Section II comprises of
problem definition & solution for the same to extract the
digital watermarked image. In section III (A) the various
Estimators to make the method robust against noisy attack.
We being discussed in Section III (B) the Statistical
parameters evaluated to check the quality of extracted
watermark image. Section IV proposes various evaluation
parameters being measured on various images & Result of
applied approach. In our last section we conclude our paper
by enhancing the result of wavelet method against noise
attack using estimator technique.
II. PROBLEM DEFINATION
As seen watermarking technique has emerge as a solution to
the problem of copying the digital content. Recently so
many watermarking schemes have been developed in the
image domain. There are different ways to watermark image
like spatial domain & using different transform in frequency
domain [4].
The basic problem of Spatial domain method is that
it could not resist the simple noisy attack while the wavelet
method of watermarking is capable to resist the noise attack
but if the image is degraded by high density noise then this
method fails to extract the actual watermark. So that is why
we have applied estimator approach to make this method
robust against noise attacks. Here we have applied M-
estimator to remove the effect of different noises at the
extraction part.
The basic flow of our algorithm is shown in below fig.1
Fig. 1: Proposed approach
Basically estimator estimates the data & fit a line which
effectively rejects the outlier & makes the system robust
against outlier. In our case the outlier is noise so that if we
apply estimator after extraction process then it estimates the
noise pixels & effectively rejects it. As a result the
perceptual quality of watermark becomes better. After
applying estimator we have evaluated our scheme using
different quality attributes. We have also compared these
results with the result of simple spatial embedding method.
Images
to
Waterma
rk
Wavelet
method
of
Waterma
rking
Apply
Noise
Attack on
Waterma
rked
Image
Extract
the
watermar
k image &
Apply
Estimator
on result](https://image.slidesharecdn.com/ijsrdv1i3017-140801061342-phpapp01/85/Comparison-of-Wavelet-Watermarking-Method-With-without-Estimator-Approach-1-320.jpg)
![Comparison of Wavelet Watermarking Method With & without Estimator Approach
(IJSRD/Vol. 1/Issue 3/2013/0017)
All rights reserved by www.ijsrd.com
469
III. MATERIALS & METHODS
In our proposed algorithm, hiding of watermark in cover
image has been done using spatial domain with & without
estimator. For that we have chosen different grey scale
images which are shown below:
(a) (b) (c)
(d) (e) (d)
Fig. 2. (a) Cell (used as a Watermark Image) and
(b) Moon, (c) Circuit, (d) Bag, (e) Cameraman, (f) Mandi
((b),(c),(d),(e) used as a Cover Image)
IV. ESTIMATOR
Estimator basically works as a filter but the advantage of
estimator is that it gives filtering result with preservation of
fine detail. Normally, the available denoizing filter blurs the
image after filtering. From the available estimators we have
used M estimator. The robustness of any estimator depends
on two parameters: Influence Function & Breakdown Point
[5]. Influence Function gives the change in an estimate
caused by insertion of outlying data and Breakdown Point is
the largest percentage of outlier data points that will not
cause a deviation in the solution. So the robustness of any
estimator depends on these two parameters. The outlier in
our case is noise attack. M estimator effectively rejects the
outlier so that we can use it to remove from the extracted
watermark without knowing the noise density. We have here
used M estimator cauchy function because it is effectively
rejects the noise. The Robustness of any estimator is defined
by the above explained two parameters. We have applied M
estimator on the extracted watermark & the given result
shows the same. To embed watermark we have used haar
wavelet transform based on the past research. Steps of the
propose algorithm for the process of watermark embedding
is mention in table1 & for extraction of the same is in table
2.In the extraction method we have applied estimator to
reject the effect of Noise. We have here described the
algorithm with Estimator approach.
Sr.
No.
Steps
1 Select one grey scale image as a Cover image.
2 Take one watermark image of same size.
3 Perform 2-level wavelet decomposition on both the
images.
4
Select the lowest frequency component of cover
Image & Highest frequency component of
watermark.
5
Reconstruct the image using above specified
component.
Table. 1: Embedding Process
Sr.
No.
Steps
1 First take the watermarked image.
2 Apply different noise attack on watermarked image.
2 Perform Wavelet Decomposition on watermarked
image.
3 Select the lowest frequency component of original
watermark & higher frequency component of
watermarked image.
4 Reconstruct the image using above specified
component.
5 Apply Estimator on extracted watermark image.
6 Calculate the statistical parameters.
Table. 2: Extraction Process
V. STATISTICAL PARAMETERS
To measure the imperceptibility & robustness of any
watermarking technique PSNR & MSE are the two major
parameters.
MSE: Mean Square Error & Root Mean Square Error is
usually used to measure perceptual quality of image. It finds
error between watermarked image and the one without
watermark
.
∑ ∑ ) )) )
PSNR: Peak Signal To Noise Ratio usually used to measure
the imperceptibility of watermarking method. It gives the
measure of invisibility of watermark in the original signal.
( ) )
This attributes are based on the objective criteria. So it is
necessary to check same technique on different objects to
measure the perfect range or value.
Correlation Coefficient: It gives the correlation
between original watermark & extracted watermark.
This attributes are based on the objective criteria. So it is
necessary to check same technique on different objects to
measure the perfect range or value.
VI. RESULTS & ANALYSIS
To check the fidelity of outcomes the proposed approach is
applied on four different grey scale cover images & the
attributes so calculated are shown in table 3. The result of](https://image.slidesharecdn.com/ijsrdv1i3017-140801061342-phpapp01/85/Comparison-of-Wavelet-Watermarking-Method-With-without-Estimator-Approach-2-320.jpg)
![Comparison of Wavelet Watermarking Method With & without Estimator Approach
(IJSRD/Vol. 1/Issue 3/2013/0017)
All rights reserved by www.ijsrd.com
471
Mandi.tif 244.0666 265.7713
Bag.tif 437.7578 355.6137
Table 3.(g) Mse based comparison
Noise with Density
Salt & pepper 0.05
Without Estimator
Salt & pepper 0.05
With Estimator
Cameraman.tif 40.1943 42.5619
Circuit.tif 47.5616 48.3653
Moon.tif 48.1839 48.4115
Mandi.tif 48.5114 47.7714
Bag.tif 43.4369 45.2420
Table 3.(b)PSNR based Comparison
Noise with Density
Salt & pepper 0.05
Without Estimator
Salt & pepper 0.05
With Estimator
Cameraman.tif 0.9783 0.9834
Circuit.tif 0.9905 0.9912
Moon.tif 0.9910 0.9914
Mandi.tif 0.9915 0.9906
Bag.tif 0.9849 0.9874
Table 3.(c)Correlation based comparison
Plot for the above tables are as shown below:
Salt & pepper noise:
MSE:
PSNR:
Correlation Coefficient:
Fig. 3: Charts for different variable and comparison
VII. CONCLUSION
From the above results we conclude that by adding the
estimator step at the extraction part gives better results
compare to without estimator. From the given result we can
observe that MSE decreases and PSNR increases. The value
of correlation coefficient shows the correlation between
extracted images with M estimator is higher compare to
simple WAVELET method.
ACKNOWLEDGEMENT
We are very thankful to Dr. Chintan Modi & Mr. Pratyaksh
Maru for providing support and their knowledge of robust
estimators. We are also thankful to our institute G.H.Patel
college of Engg. & Technology, for providing the necessary
platform for our research.
REFERENCES
[1] Frank Hartung, Martin Kutter, “Multimedia
Watermarking Techniques”, Proceedings of The IEEE,
Vol. 87, No. 7, pp. 1085 – 1103, July 1999.
[2] Ajit kulkarni, Digital watermarking”,American Journal
of computer & Multimedia, conference’04, Chicago.
[3] Mei Jiansheng, Li Sukang and Tan Xiaomei, “A Digital
Watermarking Algorithm Based On DCT and DWT” ,
International Symposium on Web Information Systems
and Applications (WISA’09) ,Nanchang, P. R. China,
ISBN 978-952-5726-00-8, May 22-24, 2009, pp. 104-
107, Proceedings of the 2009.
[4] Chandra Mohan B and Srinivas Kumar S , “Robust
Multiple Image Watermarking Scheme using Discrete
Cosine Transform with Multiple Descriptions”
International Journal of Computer Theory and
Engineering, ISSN: 1793-8201, December, 2009.
[5] Jayesh D. Chauhan, et. al., “Robust M estimator for
surface roughness estimation using Machine vision”,
International Conference on Advances in Mechanical
Engineering (ICAME), SVNIT, Surat, August 3-5, 2009.
[6] Dr.B.EswaraReddy, P. Harini, S. Maruthu Perumal &
Dr. V. Vijaya Kumar: “A New Wavelet Based Digital
Watermarking Method for Authenticated Mobile Signals
International Journal of Image Processing (IJIP), Volume
5, Issue 1, 2011.
[7] DineshKumar and Vijay Kumar, “Contourlet Transform
Based Watermarking For Colour Images” The
International Journal of Multimedia & Its Applications
(IJMA), ISSN : 0975 – 5578, February 2011.
0
200
400
600
800
MSE
Test Images
Without
Estimator
With Estimator
0
10
20
30
40
50
PSNR
Test Images
Without
Estimator
With
Estimator
0.9
0.95
1
Corr.Coef.
Test Images
Without
Estimator
With
Estimator](https://image.slidesharecdn.com/ijsrdv1i3017-140801061342-phpapp01/85/Comparison-of-Wavelet-Watermarking-Method-With-without-Estimator-Approach-4-320.jpg)
Comparison of Wavelet Watermarking Method With & without Estimator Approach
- 1. IJSRD - International Journal for Scientific Research & Development| Vol. 1, Issue 3, 2013 | ISSN (online): 2321-0613 All rights reserved by www.ijsrd.com 468 Comparison of Wavelet Watermarking Method With & without Estimator Approach Neha Y. Joshi1 Kavindra R. Jain2 Pratyaksh A Maru3 1 Research Scholar 2 Assistant Professor, G. H. Patel College of Engg. and Tech., Vallabh Vidyanagar, India 3 Assistant Professor, Dr. Jivraj Mehta Institute of Tech, Mogar, India Abstract — In this paper we propose an Estimator approach with wavelet watermarking method which is capable to hide watermark in the host image based on wavelet domain technique. Using the Estimator approach the proposed technique becomes robust against different noise attacks. For the evaluation of Imperceptibility & Robustness of the proposed method we have calculated basic statistical parameters. We have tested watermarked image against different noise attacks at different noise densities. Due to the use of estimator the perceptible quality of extracted image is better though the image is degraded by high density noise. I. INTRODUCTION Nowadays technologies are changing so to use digital watermarking instead of the age old techniques for hiding information & protecting multimedia data has become the greatest area of interest. This makes the availability of digital information in the form of audio, video & images are quiet easily and immensely available and in reach to public via web. The main concern in such techniques is the robustness & imperceptibility. The watermarking can be applied to the multimedia data like Images, text, audio & video. Watermarking technique mainly contains two processes embedding & extracting for that different types of watermark & various watermarking techniques are available. According to the properties of watermark it can be divided into two main categories visible & invisible watermark[2]. Human audio & video system means Human visual perception decides the properties of watermark. Watermarking technique can also be classified based on the extraction process. According to it, we can divide it into three types Blind, Semi-Blind & Non-Blind. Techniques which do not require the original data or signal fall into the first category which is blind. Techniques which require the original watermark are considered as a Semi-Blind watermarking & techniques which require the original signal for extraction process is the Non-Blind technique. According to the domain we can also classified watermarking as a spatial & Frequency domain. In frequency domain we cannot directly embed watermark for that first we have to convert original signal & watermark into frequency domain using different transforms while in spatial domain we can directly apply watermark by changing pixel values or by using spread spectrum approach [3]. The past researches suggest that compare to spatial domain; frequency domain watermarking is more robust against different attacks. To achieve data protection the watermarking technique should be imperceptible with high level of security i.e. watermarked image should not reveal any things about hiding information. The whole paper has been divided into V sections. Section II comprises of problem definition & solution for the same to extract the digital watermarked image. In section III (A) the various Estimators to make the method robust against noisy attack. We being discussed in Section III (B) the Statistical parameters evaluated to check the quality of extracted watermark image. Section IV proposes various evaluation parameters being measured on various images & Result of applied approach. In our last section we conclude our paper by enhancing the result of wavelet method against noise attack using estimator technique. II. PROBLEM DEFINATION As seen watermarking technique has emerge as a solution to the problem of copying the digital content. Recently so many watermarking schemes have been developed in the image domain. There are different ways to watermark image like spatial domain & using different transform in frequency domain [4]. The basic problem of Spatial domain method is that it could not resist the simple noisy attack while the wavelet method of watermarking is capable to resist the noise attack but if the image is degraded by high density noise then this method fails to extract the actual watermark. So that is why we have applied estimator approach to make this method robust against noise attacks. Here we have applied M- estimator to remove the effect of different noises at the extraction part. The basic flow of our algorithm is shown in below fig.1 Fig. 1: Proposed approach Basically estimator estimates the data & fit a line which effectively rejects the outlier & makes the system robust against outlier. In our case the outlier is noise so that if we apply estimator after extraction process then it estimates the noise pixels & effectively rejects it. As a result the perceptual quality of watermark becomes better. After applying estimator we have evaluated our scheme using different quality attributes. We have also compared these results with the result of simple spatial embedding method. Images to Waterma rk Wavelet method of Waterma rking Apply Noise Attack on Waterma rked Image Extract the watermar k image & Apply Estimator on result
- 2. Comparison of Wavelet Watermarking Method With & without Estimator Approach (IJSRD/Vol. 1/Issue 3/2013/0017) All rights reserved by www.ijsrd.com 469 III. MATERIALS & METHODS In our proposed algorithm, hiding of watermark in cover image has been done using spatial domain with & without estimator. For that we have chosen different grey scale images which are shown below: (a) (b) (c) (d) (e) (d) Fig. 2. (a) Cell (used as a Watermark Image) and (b) Moon, (c) Circuit, (d) Bag, (e) Cameraman, (f) Mandi ((b),(c),(d),(e) used as a Cover Image) IV. ESTIMATOR Estimator basically works as a filter but the advantage of estimator is that it gives filtering result with preservation of fine detail. Normally, the available denoizing filter blurs the image after filtering. From the available estimators we have used M estimator. The robustness of any estimator depends on two parameters: Influence Function & Breakdown Point [5]. Influence Function gives the change in an estimate caused by insertion of outlying data and Breakdown Point is the largest percentage of outlier data points that will not cause a deviation in the solution. So the robustness of any estimator depends on these two parameters. The outlier in our case is noise attack. M estimator effectively rejects the outlier so that we can use it to remove from the extracted watermark without knowing the noise density. We have here used M estimator cauchy function because it is effectively rejects the noise. The Robustness of any estimator is defined by the above explained two parameters. We have applied M estimator on the extracted watermark & the given result shows the same. To embed watermark we have used haar wavelet transform based on the past research. Steps of the propose algorithm for the process of watermark embedding is mention in table1 & for extraction of the same is in table 2.In the extraction method we have applied estimator to reject the effect of Noise. We have here described the algorithm with Estimator approach. Sr. No. Steps 1 Select one grey scale image as a Cover image. 2 Take one watermark image of same size. 3 Perform 2-level wavelet decomposition on both the images. 4 Select the lowest frequency component of cover Image & Highest frequency component of watermark. 5 Reconstruct the image using above specified component. Table. 1: Embedding Process Sr. No. Steps 1 First take the watermarked image. 2 Apply different noise attack on watermarked image. 2 Perform Wavelet Decomposition on watermarked image. 3 Select the lowest frequency component of original watermark & higher frequency component of watermarked image. 4 Reconstruct the image using above specified component. 5 Apply Estimator on extracted watermark image. 6 Calculate the statistical parameters. Table. 2: Extraction Process V. STATISTICAL PARAMETERS To measure the imperceptibility & robustness of any watermarking technique PSNR & MSE are the two major parameters. MSE: Mean Square Error & Root Mean Square Error is usually used to measure perceptual quality of image. It finds error between watermarked image and the one without watermark . ∑ ∑ ) )) ) PSNR: Peak Signal To Noise Ratio usually used to measure the imperceptibility of watermarking method. It gives the measure of invisibility of watermark in the original signal. ( ) ) This attributes are based on the objective criteria. So it is necessary to check same technique on different objects to measure the perfect range or value. Correlation Coefficient: It gives the correlation between original watermark & extracted watermark. This attributes are based on the objective criteria. So it is necessary to check same technique on different objects to measure the perfect range or value. VI. RESULTS & ANALYSIS To check the fidelity of outcomes the proposed approach is applied on four different grey scale cover images & the attributes so calculated are shown in table 3. The result of
- 3. Comparison of Wavelet Watermarking Method With & without Estimator Approach (IJSRD/Vol. 1/Issue 3/2013/0017) All rights reserved by www.ijsrd.com 470 watermark embedding & extraction on the images are shown below in table 4 Figure 1 shows four different grey cover images & watermark image. Figure 2 shows Images after Embedding Watermark. Cover Image Watermark Image Watermarked Image Salt & Pepper Noise(density 0.05) Extracted Watermark Without Estimator Approach Extracted Watermark With Estimator Approach Table.3 : Comparison of Extracted Watermark with & without Estimator After applying estimator we have compare the two results of applied approach by calculating different statistical parameters. We have used three different noises as attack. Table 3 Comparison Based on Statistical Parameters Salt & Pepper noise: Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 626.2221 391.6110 Circuit.tif 585.0728 675.1084 Moon.tif 701.8396 513.7373 Mandi.tif 622.4515 488.6302 Bag.tif 669.9769 281.9712 Table 3.(a) Mse based comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 40.3270 44.4045 Circuit.tif 40.9174 39.6741 Moon.tif 39.3368 42.0468 Mandi.tif 40.3795 42.4820 Bag.tif 39.7404 47.2575 Table 3.(b) PSNR based comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 0.9786 0.9863 Circuit.tif 0.9799 0.9772 Moon.tif 0.9760 0.9822 Mandi.tif 0.9787 0.9831 Bag.tif 0.9771 0.9900 Table 3.(c) Correlation coefficient based comparison Gaussian Noise: Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 1.6909e+003 386.2296 Circuit.tif 1.5501e+003 381.4452 Moon.tif 1.0345e+003 373.6339 Mandi.tif 1.5656e+003 611.6894 Bag.tif 1.5134e+003 444.7307 Table 3.(d) Mse based Comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 31.6992 44.5247 Circuit.tif 32.4543 44.6330 Moon.tif 35.9669 44.8127 Mandi.tif 32.3680 40.5310 Bag.tif 32.6627 43.2997 Table 3.(e)PSNR based comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 0.9453 0.9864 Circuit.tif 0.9495 0.9865 Moon.tif 0.9654 0.9868 Mandi.tif 0.9490 0.9792 Bag.tif 0.9506 0.9843 Table3(f) Correlation coefficient based comparison Speckle Noise Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 635.8647 484.1536 Circuit.tif 272.2698 248.2074 Moon.tif 253.4473 246.8898
- 4. Comparison of Wavelet Watermarking Method With & without Estimator Approach (IJSRD/Vol. 1/Issue 3/2013/0017) All rights reserved by www.ijsrd.com 471 Mandi.tif 244.0666 265.7713 Bag.tif 437.7578 355.6137 Table 3.(g) Mse based comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 40.1943 42.5619 Circuit.tif 47.5616 48.3653 Moon.tif 48.1839 48.4115 Mandi.tif 48.5114 47.7714 Bag.tif 43.4369 45.2420 Table 3.(b)PSNR based Comparison Noise with Density Salt & pepper 0.05 Without Estimator Salt & pepper 0.05 With Estimator Cameraman.tif 0.9783 0.9834 Circuit.tif 0.9905 0.9912 Moon.tif 0.9910 0.9914 Mandi.tif 0.9915 0.9906 Bag.tif 0.9849 0.9874 Table 3.(c)Correlation based comparison Plot for the above tables are as shown below: Salt & pepper noise: MSE: PSNR: Correlation Coefficient: Fig. 3: Charts for different variable and comparison VII. CONCLUSION From the above results we conclude that by adding the estimator step at the extraction part gives better results compare to without estimator. From the given result we can observe that MSE decreases and PSNR increases. The value of correlation coefficient shows the correlation between extracted images with M estimator is higher compare to simple WAVELET method. ACKNOWLEDGEMENT We are very thankful to Dr. Chintan Modi & Mr. Pratyaksh Maru for providing support and their knowledge of robust estimators. We are also thankful to our institute G.H.Patel college of Engg. & Technology, for providing the necessary platform for our research. REFERENCES [1] Frank Hartung, Martin Kutter, “Multimedia Watermarking Techniques”, Proceedings of The IEEE, Vol. 87, No. 7, pp. 1085 – 1103, July 1999. [2] Ajit kulkarni, Digital watermarking”,American Journal of computer & Multimedia, conference’04, Chicago. [3] Mei Jiansheng, Li Sukang and Tan Xiaomei, “A Digital Watermarking Algorithm Based On DCT and DWT” , International Symposium on Web Information Systems and Applications (WISA’09) ,Nanchang, P. R. China, ISBN 978-952-5726-00-8, May 22-24, 2009, pp. 104- 107, Proceedings of the 2009. [4] Chandra Mohan B and Srinivas Kumar S , “Robust Multiple Image Watermarking Scheme using Discrete Cosine Transform with Multiple Descriptions” International Journal of Computer Theory and Engineering, ISSN: 1793-8201, December, 2009. [5] Jayesh D. Chauhan, et. al., “Robust M estimator for surface roughness estimation using Machine vision”, International Conference on Advances in Mechanical Engineering (ICAME), SVNIT, Surat, August 3-5, 2009. [6] Dr.B.EswaraReddy, P. Harini, S. Maruthu Perumal & Dr. V. Vijaya Kumar: “A New Wavelet Based Digital Watermarking Method for Authenticated Mobile Signals International Journal of Image Processing (IJIP), Volume 5, Issue 1, 2011. [7] DineshKumar and Vijay Kumar, “Contourlet Transform Based Watermarking For Colour Images” The International Journal of Multimedia & Its Applications (IJMA), ISSN : 0975 – 5578, February 2011. 0 200 400 600 800 MSE Test Images Without Estimator With Estimator 0 10 20 30 40 50 PSNR Test Images Without Estimator With Estimator 0.9 0.95 1 Corr.Coef. Test Images Without Estimator With Estimator