A SEMI-BLIND WATERMARKING SCHEME FOR RGB IMAGE USING CURVELET TRANSFORM

International Journal in Foundations of Computer Science & Technology (IJFCST) Vol.7, No.1, January 2017
DOI:10.5121/ijfcst.2017.7101 1
A SEMI-BLIND WATERMARKING SCHEME
FOR RGB IMAGE USING CURVELET
TRANSFORM
Ranjeeta Kaushik1
, Sanjay Sharma2
and L. R. Raheja3
1
Department of Computer Science and Engineering, Thapar University, Patiala, Punjab,
India
2Department of Electronics and Communication Engineering, Thapar University, Patiala,
Punjab
3
Department of Navel Engineering,IIT, Kharagpur, India.
ABSTRACT
In this paper, a semi-blind watermarking technique of embedding the color watermark using curvelet
coefficient in RGB cover image has been proposed. The technique used the concept of HVS that the human
eyes are not much sensitive to blue color. So the blue color plane of the cover image is used as embedding
domain. A bit planes method is also used, the most significant bit (MSB) plane of watermark image is used
as embedding information. Selected scale and orientation of the curvelet coefficients of the blue channel in
the cover image has been used for embedding the watermark information. All other 0-7 bit planes are used
as a key at the time of extraction. The results of the watermarking scheme have been analyzed by different
quality assessment metric such as PSNR, Correlation Coefficient (CC) and Mean Structure Similarity Index
Measure (MSSIM). The experimental results show that the proposed technique gives the good invisibility of
watermark, quality of extracted watermark and robustness against different attacks.
KEYWORDS
Digital Watermarking, Curvelet Transform, Bit Plane, MSSIM
1. INTRODUCTION
The internet is being used by numbers of user and this number increasing day by day. Due to the
internet and the advantage of digital, Information material (multimedia data) is being distributed
without loss of quality. So it’s very difficult to protect the interest of an author or to preserve the
copyright. The most effective solution to the unauthorized distribution problem is watermarking.
Watermarking is a technique of embedding author information into the original work.
Watermarking has following requirements.
1. Watermark should not be perceptual visible
2. Watermark should be difficult to remove or impossible for modify.
3. Watermark should be robust against the image distortions caused by attackers
There are two domains for embedding the watermark one is spatial domain and other is transform
domain. In the first method of embedding, the watermark is by directly changing the original

2
pixel intensity of the cover image. The disadvantage of spatial techniques is that it does not give
the good robustness against the image processing attacks [1-2]. The second one is transformed
domain, in which watermark is embedded by changing the frequencies of the cover image. The
later one gives the good robustness. Embedding in transformations domain are common, some
famous transform of watermarking are Fourier transform [3], Discrete Cosine Transform (DCT)
[4-8]; Digital Wavelet Transform (DCT) [9-13], Ridgelet Transform (RT) [14] and much more.
Cox et al. [2] renowned that watermark to be robust against the image processing operations, it
must be located in perceptually important regions of the image. Watermarking is based on
samples and embedded into largest DCT coefficients. R. Dugad [10] &Peining Tao [11] proposed
watermarking using wavelet transformation which provides the multi-resolution representation of
the image. J.-J. Lee [12] proposed a dual-tree implementation for complex wavelet transform
(DT-CWT), which attains good directional selectivity and approximate shift invariance.
2. LITERATURE REVIEW
In literature, many authors embedded the watermark in the wavelet domain. Though wavelet
transform has been explored broadly in image processing, due to the problem of representing line
singularity, it fails to represent edges and curves [16-18]. The wavelet transform is not suitable
for describing anisotropic elements. It includes only two directional elements it means transform
is independent of scale. The disadvantage is overcome by the ridgelet transform. The basic idea
of the ridgelet transform [14] is to plot a line singularity in the 2-D domain into a point by means
of a radon transform. There is also some watermarking based on the ridgelet transform. But
ridgelet transform cannot well represent curve edges. So Curvelet transform is developed, it is a
multiscale pyramid with many directions and positions at each fine scale and has needle-shaped
elements at the fine scales [15-16].
Researchers have changed their focus to the curvelet transform [18] for its properties such as it
provide sparse representations of objects along a curve and it is anisotropic with strong direction.
Thai Hien et al. [19] embedded the watermark in the curvelet transform which contains as much
edge information as possible. It has good invisibility but poor robustness. The method of Thai
Hien et al. [20] embedded a watermark in curvelet coefficients which are selected by a threshold.
This method has good invisibility and robustness. Shi et al. [21] proposed a semi-fragile
watermarking algorithm by embedding the watermark in the supreme module of curvelet
coefficient. This technique gives good robustness against compression operation.Among these
techniques, those require the original cover image, key and bits of watermark for watermark
extraction are called non-blind watermarking techniques [4]. Robustness of non-blind technique is
good under image processing attacks. But it is not appropriate for watermark detection in DVD
player because the original data is not accessible. Those need the secret key but not require
original cover image at the time of extraction is called a blind watermarking technique. This
technique is suitable for all types of application, but the blind watermarking is usually less robust
compared with the non-blind watermarking scheme. Those require the key and the watermark bit
sequence is called semi-blind watermarking technique. Semi-blind watermarking techniques give
the yield good robust against attacks and suitable for most of the application. In this paper, we
discuss the semi-blind technique of watermarking.Literature also gives some semi-blind
technique Lin et al. [22] proposed a semi-blind watermarking scheme using the Discrete Fourier
Transform (DFT). Solachidis et al [23] use circular symmetric in DFT domain. Licks et al. [24]
proposed a circular symmetric watermarking in which watermark is extracted by an exhaustive
search. Stankovic et al. [25] embed the watermark by mean of a 2D radon –Wigner distribution.

3
There is not so much literature found on color watermarking in curvelet domain. Some of the
author takes the colored image as the cover image in best of author knowledge no one tried to
embed the color watermark in the color cover image using curvelet transform. In this paper, we
tried to embed an RGB watermark image into curvelet coefficients of RGB cover image. Here
author used a bit plane method to split watermark image. The Most significant bit is used an
embedding information and other lower bit planes used as the key. To evaluate the invisibility
and the robustness of proposed method PSNR, Normalized Correlation (NC) and MSSIM metrics
are used.
3. CURVELET TRANSFORM
This block ridgelet-based transform is called as curvelet transform. It was first proposed by
Candes and Donoho [15]. Actually the ridgelet transform is the fundamental for the curvelet
transform. The ridgelet transform is optimal at representing straight-line singularities.
Unfortunately, worldwide straight-line singularities are rarely used in many applications [31].
To analyze local line or curve singularities, the natural idea is to consider a partition of the image,
and then to apply the ridgelet transform to the obtained sub-images. So the block based ridgelet
transform is named as curvelet transform[31-32].
Let µ be the triple in the function plane. Where j=0,1…. is a scale parameter, l= 0,1,2…..
an orientation parameter and is the translation parameter pairs. A
curvelet coefficient is simply the inner product between an element and curvelet
given by
Where is the rotation by radian, J= (j,l) is the index of wedge for all k with it and is a
polar wedge window of radial dilation and angular transformation[17]. Define coarse scale
curvelet as
Similar to other multi-scale pyramids, curvelet transform images into several frequency scales.
An important factor in curvelet domain is not the “Approximation Rate”. If having an object in
the domain , how fast can approximate it using the certain system of function?

4
Eq. (iv), Eq. (v) and Eq. (vi) gives the approximate rate of Fourier Transform, wavelet transform
and curvelet transform respectively. Conceptually the curvelet transform is a multiscalepyramid
with many directions and positions at each fine scale and needle-shaped elements at fine scales
[30]. Based on the curvelet transform, it is possible to embed watermarks onto more significant
components and spread the amendments to more space locations.
4. PROPOSED WATERMARK EMBEDDING TECHNIQUE
Here we proposed a technique of embedding the color watermark in an RGB image. To embed
the watermark, the original cover image is being divided into color planes. The cover image has
three color planes red, green and blue. The embedding has following steps
4.1 Embedding algorithm
a) Watermark image is also colored so firstly separates there color planes Br, Bg and Bb red
plane, green plane and blue plane of the watermarked image respectively. Now obtain bit
planes of each color plane. {Br1, Br2….Br8}, {Bg1, Bg2…Bg8} and {Bb1, Bb2…Bb8} are the set
of bit planes of red, green and blue color planes respectively. Br1.Bg1 and Bb1 is the least
significant bit of red, green and blue color. Br8, Bg8 and Bb8 are most significant bit plane
of red, green and blue color used as the watermark (to be embedded) and remaining other
{Br1, Br2….Br7}, {Bg1, Bg2…Bg7} and {Bb1, Bb2…Bb7} are used as a key at the time of
extraction.
b) The frequency spectrum is independent of transmission medium. Absorption spectrum
gives the idea about the spectral sensitivity of 3 cones. This shows that the eyes are not
much sensitive to blue color. The processed watermark has been embedded into the blue
color plane of frequency domain of the cover image. The semi blind method of
embedding watermark in curvelet domain is given below
c) Read the original cover image C and decomposed cover image C into RGB color planes.
Let Cr, Cg and Cb represent the red, green and blue color plane of the original cover image
respectively.
d) Perform a curvelet transform on the blue color plane (Cb) with a j number of
decomposition level and l number of scales. In this paper 4 decomposition level and 3
scales are used. The resultant curvelet coefficients being stored in multidimensional array
Clet
e) Select the orientation and scale for embedding the watermark. Let selected scale j and
orientation k1, k2 and k3 for embedding the bit planes Br8, Bg8 and Bb8 respectively. Stored
the selected oriented and scale coefficients
f) Set α i.e the key of embedding.
g) For each coefficients of selected orientation and scale modified by
If
• Else

5
• Else
Else
h) Apply inverse curvelet transform with the same scale on to the modified coefficients to
change the resulted image from frequency domain to time domain. Add this modified
blue color plane to red and green plane, means to convert resulting image into an RGB
watermarked image
4.2 Extraction Algorithm
Read watermarked image and obtains its color plane Wr, Wg and Wb are the color planes of red,
green and blue color respectively. Apply selected scale and perform curvelet transform on blue
color plane (Wb) of watermarked image. Stored the curvelet coefficient let Ex
a) Apply selected scale and orientation where watermark had been embedded. Let selected
scale j and orientation k1, k2,k3 for red, green and blue bit planes
b)
c) For each coefficient of ESr, ESg, ESb
d) Extract each bit planes of watermark
a.
b.
c.
e) From the 6th
step we find the most significant bit planes now we used stored watermark
bit plane to extract the color watermark. For each color plane and for each bit.
Extr=EWr *2^7+ Br7*2^6+ Br6*2^5+ Br5*2^4+Br4*2^3+Br3*2^2+Br2*2^1+Br1*2^0;
Extg=EWg*2^7+ Bg7*2^6+ Bg6*2^5+ Bg5*2^4+Bg4*2^3+Bg3*2^2+Bg2*2^1+Bg1*2^0;
Extb=EWb*2^7+ Bb7*2^6+ Bb6*2^5+ Bb5*2^4+Bb4*2^3+Bb3*2^2+Bb2*2^1+Bb1*2^0;
f) Extr, Extg and Extb are the color planes of red, green and blue color plane of extracted
watermark image. Combine these color planes and get the RGB extracted watermark.
5. QUALITY ASSESSMENT METRICS
Embedding the watermark in the cover image may degrade visual quality of the image. Quality
assessment metrics are based on well mathematical models that can predict visual quality by
comparing a watermarked image against a cover image. Following are the various metrics used
for the measurement of visual quality comparison.
5.1 Peak Signal to Noise Ratio (PSNR)
The most common evaluation method is to compute the peak signal-to-noise ratio (PSNR)
between the host and watermarked signals. PSNR is defined as follows:

6
More PSNR means better image quality. Where MSE (Mean Square Error) is define as
2
Where Im = Original image, Iw= watermarked image
5.2 Correlation Coefficient (CC)
Correlation Coefficient is the well knows measure for comparing the visual quality of two
images. The metric of CC defined as follows.
,
where , are the pixel intensities of cover and watermarked image respectively with m*n
size.
Although CC and PSNR is convenient to calculate and also has clear physical meaning but still it
does not correlate strongly enough with the visual quality of the image for most applications
[26].The scope of perceptual quality is widened in the Mean Structure Similarity Index Measure
(MSSIM) [27-29].
5.3 Mean Structure Similarity Index Measure (MSSIM)
MSSIM is the combination of three perceptual properties namely Luminance, Contrast and
structure at every point of the two images being compared. For the purpose of calculation of
MSSIM the image to be compared are broken in to number of windows of G*G, further the two
images must be of the same size [27]. If x and y are two images, M is the number of G*G
windows and N is the number of pixels in each window then MSSIM is define in the following
steps.
SSIM = ,
where ,
,
, In paper [28] k1= 0.01, k2=0.03 has been proposed
L= Dynamic range of the pixel values (255 for 8 bit gray scale image)
MSSIM (Mean SSIM) =
6. EXPERIMENTAL RESULTS AND DISCUSSION
To test the algorithm colored “lana.jpg” image used a cover image and a colored “thaper.jpg”
image used a watermark image. Fig. 1 and Fig.2 show the “lana.jpg” and “thaper.jpg” image

7
respectively. Fig. 3 shows the resulting watermarked image. Algorithm has been tested against all
the requirement parameter such as invisibility, effectiveness, and robustness. Here we compared
the quality of watermarked image and extracted watermark by the above-defined quality
assessment metrics such as PSNR, CC, and MSSIM. To show that the algorithm gives the good
robustness, the different image processing operations are being applied to the watermarked image.
The quality of extracted watermark from the distorted image have been analyzed by the quality
assessment metrics.
6.1 Invisibility
Embedding extra information in the original signal will cause degradation and perceptual
distortion. [4]. the most common evaluation method is to compute the peak signal-to-noise ratio
(PSNR) between the cover image and watermarked image. In this paper to test the invisibility of
the algorithm the quality of watermarked image and cover image are compared with the help of
quality assessment metrics. Fig 1.3 shows the watermarked image.The quality of watermarked
image has been analyzed by PSNR, CC and MSSIM metrics on each plane. Table 1.1 shows the
values of quality assessment metrics. In Table 1.1 Red_CC, Gr_CC, Blue_CC, Red_MSSIM,
Gr_MSSIM and Blue_MSSIM represent the correlation coefficients (CC) and MSSIM of the red,
green and blue color respectively. It may be observed that the PSNR is 44.97 confirming a high
invisibility of the watermark. Besides, the correlation coefficient, MSSIM is 0.99 for image
confirming almost no difference between the cover and watermarked image. This clearly shows
that the watermark is perfectly invisible.
Table 1.1 Invisibility Test
Image Name PSNR Red_CC Gr_CC Blue_CC Red_MSSIM Gr_MSSIM Blue_MSSIM
Watermarked 44.974 0.9996 0.9996 0.9991 0.9941 0.9950 0.9943
Table 1.2 Effectiveness Test
Extracted
watermark
Image
PSNR Red_C
C
Gr_CC Blue_CC Red_MSSIM Gr_MSSIM Blue_MSSIM
Fig. 4 38.5623 1 1 1 0.9761 0.9981 0.9953
Fig. 4 Extracted watermarkFig. 3 Watermarked ImageFig. 2 thapar.jpg
Fig. 1 lana.jpg

8
6.2 Extraction
Effectiveness refers to whether it is possible to detect a watermark immediately following the
embedding process. To test the effectiveness of the algorithm the same but in the reverse scheme
is used to extract the watermark from the watermarked image. There is no need of cover image
while extracting the watermark because the above discuss algorithm gives the semi-blind
extraction. Fig. 4 shows the extracted watermark image. The extracted watermark is compared
with the original watermark. Table 1.2 shows the extracted watermark and the quality of extracted
watermark is being analyzed by quality assessment metrics. The PSNR of extracted watermark is
38.5623, CC is one in all planes and MSSIM is also quite good. Table 1.2 verifies that the visual
quality of extracted watermark is very good and highly matched with original one.
6.3 Robustness Test
Robustness refers to the ability of the detector to detect the watermark after signal distortion, such
as format conversion, the introduction of transmission channel noise and distortion due to channel
gains. In order to test the robustness of the technique, one requires adding some sort of noise into
the watermarked image and then extracting the watermark. Besides, robustness is also tested by
applying various image processing operations such as filtering, rotation, cropping, adding sparsity
and shearing on the watermarked image and then extracts the watermark. Thereafter, the
extracted watermark is compared by original watermark image by applying the quality assessment
metrics. Fig. 5 to Fig. 11 show the quality of extracted watermarks from the distorted
watermarked images. Table 1.3 shows the results of extracted watermark under various attacks.
Robustness has been testing under approximately all the image processing attacks.
Fig. 5 Extracted watermark
from Gaussian Noised
watermarked image
Fig. 6 Extracted watermark
from Pepper salt noised
watermarked image
Fig. 7 Extracted watermark from
900
rotated watermarked image
Fig.8 Extracted watermark from un-
sharp filtered watermarked image
Fig. 10 Extracted watermark after projective
shearing operation on watermarked image
Fig. 11 Extracted watermark after sparsity
(128*256) on watermarked image
Fig. 9 Extracted watermark from
128*128 cropped watermarked image

9
6.3.1. Anti- Noise
Test the robustness against noise operation. Authors add two types of noise one is Gaussian noise
with 0.1 variance and other is pepper salt noise with noise density 0.01 on watermarked image
then tried to extract the watermark from noised watermarked image. Fig. 5 and Fig. 6 show the
visual quality of Extracted watermark from Gaussian Noised watermarked image and pepper salt
noised watermarked image respectively. The results are presented in Table 1.3 with two types of
noise namely pepper salt and Gaussian. The similarity measure MSSIM and CC confirm the
presence of watermark to the extent of more than 70 % similarity on red and green color planes.
The red color planes are more sensitive to human eyes.The CC and MSSIM of red color planes of
extracted watermark are 0.9. This shows that the technique is robust enough against noise
addition.
Table 1.3 Robustness test against Noise
Attacks PSNR Red_CC Gr_CC Blue_C
C
Red_MSSIM Gr_MSSIM Blue_MSSIM
Gaussian
Noise
19.631 0.9596 0.8979 0.7136 0.8998 0.7916 0.6555
Pepper
Salt Noise
18.953 0.9496 0.8579 0.7086 0.8783 0.7716 0.5955
6.3.2. Anti-Rotation
Rotation is very important to image processing operation, but most of the authors do not test the
robustness of their techniques against the rotation. Here we test the robustness against rotation
operation the watermarked image is rotated by 90 degrees and the watermark is extracted. Fig. 7
shows the visual quality of the Extracted watermark from 900 rotated watermarked image. The
extracted watermark is being compared by original watermark image using different quality
assessment metrics on each color planes. Table 1.4 shows the values of visual quality of
extracted watermark with the original one. It may be observed that both similarities measure
MSSIM and CC are reasonably good in each color planes that confirming the similarity of
extracted watermark with an original one. Therefore, it may be concluded that the technique is
robust against rotation operation.
6.3.3. Anti – Filtering
In order to check the robustness against filtering operation, the unsharp filter was applied on
watermarked image and thereafter the watermark was extracted from the filtered watermarked
image and compared with original watermark image. Fig.8 shows the visual quality of extracted
watermark from un-sharp filtered watermarked image. Table 1.4 shows the values of quality
assessment metric by comparing extracted watermark from the filtered watermark image and the
original watermark. The CC and MSSIM are good in all the color planes that show the similarity
of extracted watermark with the original one. Table 1.4 verifies that this watermarking technique
is robust against the filtering operation of image processing.

10
6.3.4 Anti- cropping
Cropping of an image means to cut a part of the image. Accordingly, it is now required to check
whether the watermark is present in the cropped part of the image. Here, cropped parts of size
128*128 of the watermarked image are taken and then the watermark is being extracted by the
cropped part of watermarked image. Fig. 9 shows the visual quality of extracted watermark. The
values of quality assessment metrics are shown in Table 1.4. It is observed that the values of
similarity metrics MSSIM, CC are slightly lower than 0.6 in general. This is obviously due to the
relatively smaller size of the cropped part. However, the values confirm the presence of the
watermark in these cropped sizes
Table 1.4 Robustness test against image processing Operations
Attacks PSN
R
Red_C
C
Gr_C
C
Blue_C
C
Red_MSSI
M
Gr_MSSI
M
Blue_MSSI
M
90 Degree
Rotation
14.64
9
0.9293 0.7614 0.6682 0.8629 0.7311 0.6216
Un-sharp
Filtering
11.23
2
0.8787 0.7583 0.7025 0.8073 0.6889 0.6665
Cropping
128*128
13.08
9
0.7843 0.6339 0.5994 0.7723 0.6047 0.5418
Projective
Shearing
8.584
5
0.8555 0.7001 0.6879 0.7858 0.6530 0.6338
Sparsity
(128*256)
11.64
9
0.8379 0.7487 0.8122 0.7912 0.6939 0.7979
6.3.5 Anti – with other attacks
To test the robustness of the technique two more attacks are performed on watermarked image
one is projective shearing and other is adding some sparsity. Sparsity means adding a number of
zeros in the image. Here a vertical 128*256 sparsity is added into watermarked image. Fig. 10
and Fig. 11 show the visual quality of extracted watermark after the projective shearing operation
and after adding 128*256 sparsity on watermarked image respectively. Table 1.4 shows the
quality of extracted watermark image from these operations. The values of CC and MSSIM in
each color planes are nearly 0.7 in general that is reasonably good confirming the similarity of
watermarks. Therefore, it may be concluded that the technique is robust against any image
processing operations.
7. CONCLUSION
The semi-blind color watermarking technique of embedding the color watermark in RGB image
has been presented. The extraction does not need the original cover image at the time of
extraction. The bit plane and HVS concepts are used for embedding the watermark. The most
significant bit planes of each color planes of watermark image are embedded into the curvelet
coefficients of the blue color plane of the original cover image. To test the scheme, the algorithm

11
has been tested against invisibility, effectiveness, and robustness. The quality of watermarked
image and extracted watermark are being analyzed by the quality assessment metric such as
PSNR, CC, and MSSIM. By the Experimental results show the above discussed watermarking
technique fulfill all the requirements of watermarking and gives a robust color watermarking
without the loss of quality of the original cover image.
REFERENCES
[1] T. S. Chen, C. C. Chang, and M.S. Hwang,A virtual image cryptosystem based upon vector
quantization, IEEE Transactions on Image Processing, 7 (1998) 1485-1488.
[2 ]C. T. Hsu and J. L. Wu , Hidden digita1 watermarks in images, IEEE Transactions on Images
Processing, 8 (1999) 58-68.
[3] C. Zhang, L. L. Cheng, Z. Qiu and L. L. Chang,Multipurpose Watermarking Based on
multiscaleCurvelet Transform, IEEE transactions on information forensics and security, 3 (2008)
611-619.
[4] I. Cox, J. Kilian, F. Leighton, and T. Shamoon, Secure Spread Spectrum Watermarking for
Multimedia, IEEE Transactions on Image Processing, 6 (1997) 1673-1687.
[5] A. Piva, M. Barni, F. Bartonlini, V. Cappellini, DCT-base watermark recovering without resorting to
the uncorrupted original image,Proc. ICIP, Atlanta, GA, 1(1997) 520–523.
[6] N. Ahmidi, R. Sfabakhsh, A novel DCT-based approach for secure color image watermarking,Proc.
Int. Conf. Information Technology: Coding and Computing, Las Vegas,(2004) 709–713.
[7] J. Huang, Y.Q.Shi, Y.Shi, Embedding image watermarks in Dc components, IEEE trans. Circuits
Syst. Video Technol.10(2000) 974-979.
[8] C. S LU, S. K. Huang, C. J. Sze, H.Y.M. Liao, Cocktail watermarking for digital image protection,
IEEE Trans. Multimedia,2 (2000) 209-224
[9] S. H. Wang, Y. P. Lin, Wavelet tree quantization for copyright protection watermarking,IEEE Trans.
Image Process., 13 (2004) 154–165.
[10] R. Dugad, K. Ratakonda, N. Ahuja, A New Wavelet-Based Scheme for Watermarking Images, Proc.
of Int. Conf. on Image Processing (ICIP 1998), Chicago, 2 (1998) 419-423,
[11] P. Tao, A. M. Eskicioglu, A Robust Multiple Watermarking Scheme in the Discrete Wavelet
Transform Domain, Optics East Internet Multimedia Management Systems Conference, Philadelphia,
PA, 2004
[12] J.J. Lee, W. Kim, N.Y.Lee, G.Y. Kim, A new incremental watermarking based on dual-tree complex
wavelet transform, J. Supercomput., 33 (2005) 133–140.
[13] Hsu, Hidden digital watermarks in images, IEEE Trans. Image Processing, 8 (1999) 55-68.
[14] M. N. Do, M. Vetterli, The finite ridgelet transform for image representation, IEEE Trans. Image
Process., 12 (2003) ,16–28.
[15] E. J. Candes and D. L. Donoho, C. R. A. Cohen , L. L. Schumaker, Eds., Curvelets - A surprisingly
effective nonadaptive representation for objects with edges, in Curves Surfaces, Nashville, TN, pp.
105–120,2000.
[16] E. J. Candes, L. Demanet, D. L. Donoho, L. Ying. , Fast discrete curvelet transforms,Multiscale
Model. Simul. 5 (2005) 861-899.
[17] E. J. Candes and D. L. Donoho, New tight frames of curvelets and optimal representations of objects
with C2 singularities, Comm. Pure Appl. Math, 57 (2004)219–266.
[18] E. J. Candes, D. L. Donoho, Continuous curvelet transform: II. Discretization and frames, Appl.
Comput. Harmon. Anal.19 (2003) 198-222.
[19] T. D. Hien, K. Miyara, et al., Curvelet Transform Based Logo Watermarking, Innovative Algorithms
and Techniques in Automation, Industrial Electronics and Tele-communication, (2007) 305-309.
[20] T. D. Hien, I. Kei, H. Harak, et al., Curvelet-Domain Image Watermarking Based on Edge-
Embedding, Lecture Notes in Computer Science, 4693 (2007) 311-317.
[21] J. P. Shi, Z. J. Zhai, Curvelet Transform for Image Authentication, Rough Set and Knowledge
Technology,(2006), 659-664.

12
[22] C.Y. Lin, M. Wu, J.A. Bloom, I.J.Cox, M. L.Miller , Y. M. Lui, Rotation scale, and translation
resilient watermarking for images, IEEE Trans. Image Processing, 10 (2001) 767-782.
[23] V.Solachidis, I. Pitas, Circularly symmetric watermarking embedding in 2 D DFT domain, IEEE
Trans. Image Processing, 10 (2001) 1741-1753.
[24] V. Lick, R. Jordan, on digital image watermarking robust to geometric transformation, Proc. IEEE Int.
conf. Image processing, 3(2001) 600-693.
[25] S. Stankovic, I. Djurovic, I. Pitas, watermarking in the space/spatial frequency domain using two
dimensional Radon-Wigner distribution, IEEE Trans. Image Processing, 10 (2001) 650-658.
[26] E. Marini, F. Autrusseau, P. Callet, P. Campisi, Evaluation of standard watermarking techniques,
Electronic imaging, security, steganography and watermarking of multimedia contents, San, 2008
[27] Z. Wangand,A. C. Bovik, A Universal Image Quality Index, IEEE signal processing letters, 9 (2002)
81-84.
[28] Z. Wang, A. C.Bovik, Image Quality Assessment: From Error Visibility to Structural Similarity,
IEEE transaction on image processing, 13 (2004) 600-612.
[29] Z. Wang and A. C. Bovik, "Mean Squared Error: Love It or Leave It?,IEEE Signal Processing
Magazine, 26 (2009)98-117.
[30] D. L. Donoho, M. R Duncan, “Digital Curvelet Transform: Strategy, Implementation and
Experiments,Proc. SPIE, Wavelet Applications VII, 4056 (2000)
[31] S. Jean-Luc, E.J.Candes, D.L.Donoho, "The curvelet transform for image denoising, Image
Processing IEEE Transactions,11 (2002) 670-684.
AUTHORS
Ms. Ranjeeta, working as Associate professor in CGC, Landran, Mohali, Punjab and
pursuing Ph.D from computer department, Thapar University, Patiala. And. The area of her
research is image processing.
Dr. Sanjay Sharma, Professor and head of department, ECE Department, Thapar University,
Patiala. He has a rich experience In the field of antenna and image processing. He has
numbers of research publications in the same field.
Dr. L.R.Raheja, Ex professor, IIT, Kharagpur. He has 40 years of teaching and research
experience. The area of expertise is naval designing, Graphics, Simulation and mathematics.

A SEMI-BLIND WATERMARKING SCHEME FOR RGB IMAGE USING CURVELET TRANSFORM

More Related Content

What's hot (17)

Viewers also liked (20)

Similar to A SEMI-BLIND WATERMARKING SCHEME FOR RGB IMAGE USING CURVELET TRANSFORM (20)

More from ijfcstjournal (20)

Recently uploaded (20)

A SEMI-BLIND WATERMARKING SCHEME FOR RGB IMAGE USING CURVELET TRANSFORM