ISSN (Print) : 0974-6846

Indian Journal of Science and Technology, Vol 9(33), DOI: 10.17485/ijst/2016/v9i33/88449, September 2016 ISSN (Online) : 0974-5645

Analysis of Current FPGA-Based Image

Watermarking Attempts
Taha Basheer Taha*, Ehkan Phaklen and Ngadiran Ruzelita
School of Computer and Communication Engineering, University Malaysia Perlis, 02060 Arau, Perlis, Malaysia;
[email protected], [email protected], [email protected]

Abstract
Objective: Analysing of current FPGA-based image watermarking attempts to figure out the future requirements for such
systems. Analysis: Hardware-based watermarking systems have been studied according to their performance in time,
frequency and robustness measurements. A comparison has been made among current systems in terms of robustness,
image quality achieved and system performance. Findings: Image watermarking was a very demand area of research in
last decades, but hardware based attempts of image watermarking are still few in compare with software implementations
although many modern systems nowadays rely on hardware-based real time algorithms. For their flexibility and
reprogrammable nature, Field Programmable Gate Arrays (FPGAs) where adopted by many researchers for implementing
watermarking hardware algorithms in cases where the need of high speed computations is a critical requirement. We
find that many current watermarking schemes need to combine the required watermarking features as image quality
measurements, using real size images, adaptable designs, and using FPGA features in using parallel high performance
systems.

Keywords: FPGA, Image Processing, Spatial domain, Transform domain, Watermarking

1. Introduction Watermarking process can be performed using

s­ oftware applications which are easier to implement but
Copyright protection for digital media is one of the most they are done using offline systems, or it can be per-
important challenges in these days. Since a rate of 70 per- formed using hardware devices in real time manner
cent of transmitted data represented by digital images are for hardware-based design, application specific inte-
critical parts of network exchange1, finding a way to pro- grated circuit (ASIC) or field programmable gate arrays
tect the copyrights of these images was a very demanding (FPGAs) can be used. The choice among them is decided
challenge for researchers and developers in last decades. after comparison in power, cost and performance5. In last
Watermarking is the process of embedding a piece of decades FPGA devices were the choice for many develop-
data like proprietary information inside digital media for ers for their ability to design almost any type of digital
copyright protection2, alteration discovery or even to check systems, reconfigurable nature, and their high perfor-
the Quality of Service (QoS) of transmitting channels3. mance since the latest FPGAs could pass the 500 MHz
The word “watermark” is referred to the data that repre- operating frequency6. Hence, in this survey, FPGA-Based
sent proprietary information, it may be called as robust image watermarking attempts are presented, and analy-
watermark if it resists and still stable while exposing to dif- sed last published watermarking approaches. Finally a set
ferent modifications. The term fragile watermark is used of future needs for such designs are listed.
when the watermark is very sensitive to any modifications, This paper is organized as following: section 2 explains
or semi fragile watermark which is a level between fragile general watermarking process, section 3 is watermarking
and robust4. Watermark can be visible as TV logos or invis- requirements, Section 4 is watermarking applications
ible which is hidden within multimedia original data. while section 5 is the review for current watermarking

*Author for correspondence

 Analysis of Current FPGA-Based Image Watermarking Attempts

techniques. Section 6 is analysing current works and watermarked image and/or secret key and/or original
list future requirements. Last but not least, section 7 is a image7. The watermarking algorithm which uses the orig-
­conclusion of survey. inal image to extract the watermark is called non blind,
while blind watermarking relies on watermarked image
only8.
2.  General Image Marking Process
Image watermarking is the process of inserting watermark 3.  Watermark Requirements
pixels to original (host) image in a way that is still hidden
(for invisible watermarking schemes) while in the same Before designing or developing a watermark system, the
time, it has a security by which only authorized people requirements of a system must be studied as the impor-
can extract it. The process of adding watermark pixels to tance of each requirement shapes the system design and
the host image is called embedding scheme (Figure 1), its methodology. The five common watermarking require-
and on the other hand the process of extracting hidden ments are as follow9
watermark pixels from the watermarked image is called i. Robustness
extracting process (Figure 2)7. Robustness is the ability to detect the watermark after
Inputs for the watermarking embedding scheme are unintended or intended distortion on watermarked data.
original image, watermark and optionally the secret key, However, this requirement is desirable in robust water-
which is used to enforce security in watermarking system. marking systems only. Since in fragile watermarking
A watermark may be an image, text or sequence number. systems with authentication duty, any data modification
Embedding scheme produces watermarked image as an is detected by the watermark damage. Hence, no need for
output. The decoding or extracting scheme inputs are robustness in such systems.
ii. Invisibility
Invisibility is to hide the watermark in a way that it can-
not be observed by Human Visual System (HVS). This
requirement is essential for invisible watermarking sys-
tems where the watermark is not preferred to be visible on
host data. A similar word for invisibility is fidelity, which
is the perceptual similarity between the host and water-
marked data.
iii.  Data Payload
Data payload is the number of embedded watermark
bitswithin the original data. That number should uniquely
Figure 1.  Watermark embedding scheme. definesthe copyright of the image. Possible attacks and
distortions must be taken into consideration while choos-
ing the data payload.
iv. Simplicity
Watermarking system must be as simple as possible,
especially while working on real time systems, where the
hardware resources are limited and the required process-
ing time is short.
v. Real-time
Although this requirement is not mentioned frequently,
but it is crucial feature in applications where no delay
is permitted between capturing data and watermark
embedding. Example of these are real time broadcasting
Figure 2.  Watermark extracting scheme. and camera system in courtroom evidence10. Hardware

2 Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology
 Taha Basheer Taha, Ehkan Phaklen and Ngadiran Ruzelita

i­mplementation is needed to apply this feature, and that 5. FPGA-Based Image

is one of the major reasons to design and implement
­hardware based watermarking algorithms.
Watermarking Survey
Current watermarking techniques could be classified
into two main classes according to the domain where
4.  Watermark Applications the watermark is embedded14. These include time spatial
Watermark systems have wide distribution nowadays, five domain and frequency transform domain. In time spa-
of major watermarking applications are listed as follow tial domain technique, watermark bits are embedded and
extracted directly to/from the image pixels. In general, it
i.  Content Identification and Copyright Protection is easy and fast for their few required resources. On the
Digital watermarking enables content of identifications other hand, frequency domain deals with transformed
asmetadata embedded in a way that it persists with media components of the image while inserting watermark
content. The watermark can carry information such as, pixels, it requires more resources and more compli-
who owns the data, date of creation, how it used, or any cated operations than time partial domain technique.
other details which owner aim to embed. Copyrights can As a result, the watermark survives against many image
be proved by extracting the watermark in type of text, ­processing manipulations14,15.
image or unique sequence4. Next subsections are spatial domain, transform
ii. Broadcast Monitoring domain, and dual mode (which is a mix between time and
In broadcasting, watermark bits are embedded within frequency) FPGA-based watermarking attempts.
shared media in a way that any attempt to remove or alter i.  Spatial Domain Attempts
the watermark will ruin the entire material where that Spatial time domain systems are easier to implement and
watermark is embedded11. Another usage of watermark- achieve real time requirement, but more design efforts
ing in broadcast monitoring is media tracking, where
are needed to obtain security and robustness. Two of
television, radio or online channels are monitored and
earliest spatial digital image watermarking architectures
when the owner’s watermark is captured from unknown
attempts using FPGAs were the schemes presented by16
resources, details for these resources are captured and
and5. 16Added the watermark which is generated using
instantly reported.
Linear Feedback Shift Registers (LFSR) to the host image,
iii.  Tamper Detection the overall encoder period was 19.842 ns with 838 cells.
Watermark is used to ensure the integrity of the data Second attempt5 was in 2007 proposed a new system
by embedding a fragile watermark in a way that it will which serve both robust and fragile watermark algo-
instantlybe destroyed if the original data is exposed to any rithms by adding the watermark to the third slice plane
tempering attempt. In semi fragile watermarking schemes, while they utilized FPGA and custom IC separately for
a watermark is still robust for some permitted modifica- this work. Both attempts have the same results for a time
tions, but it will be destroyed for malicious attacks. For and number of slices, however the quality measurements
example, image compression can be done without a­ ffecting such as PSNR is from a range of (21-56) dB by using
the watermark, but pixel alteration will ruin it12. Matlab where attacks are measured but no numeric values
for geometric ones.
iv.  Intellectual Property (IP) Protection
Two other attempts are followed by17,18. In17, proposed
Digital system designs are the result of knowledge and
a spatial semi-fragile watermarking scheme by taking
developers’ efforts, hence design methodologies that
advantage of spread spectrum and channel coding usage.
allow the reuse of IP needs watermarking techniques to
A spread spectrum is used to distribute watermark bits
protect owners and programmers rights13.
on 8*8 image blocks and channel coding was utilized
v- Quality of Service (QoS) to increase robustness. The design required 11536 slices
One of the non-conventional watermarking application is from Xilinx Virtex xc4vlx25-10ff676 FPGA to work
the QoS measurements for digital media, by evaluating within 219.542 MHz of frequency. Although encoding
the transmitting channel condition from the status of the loss caused some distortion on the decoded watermark,
received fragile watermark3. algorithm is resistant to many attacks such as filtering,

Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology 3
 Analysis of Current FPGA-Based Image Watermarking Attempts

noise, image sharpening, and compression. The next ii.  Transform Domain Attempts
attempt by18 was done using a spread spectrum utilizing a In general, transform or frequency based watermarking
binary watermark. The conversion from integer to binary schemes are resisting to some attacks and immune too
is done by Matlab. The size of the watermark was (64*64) many alterations20, but the disadvantage of such systems is
while the cover image was in (256*256). The required the complexity and high resource consuming. Developers
number of slices was 959 with a frequency of 82.26MHz. usually try to find the possible ways to economize in their
Detailed measurements of PSNR, MSSIM, normalized resources as floating number redemption or pipelining.
correlation, and mean absolute error are given for the One of the earliest FPGA-based frequency watermark-
extracted watermark but not for the original image. The ing attempts was done by21 when they presented a robust
using of real size host image and watermark is a good step image watermarking scheme on image and video by using
in watermarking, since visible and invisible watermarking Altera Stratix II FPGA. They utilized Discrete Wavelet
techniques are required to clearly express authentication Transform (DWT) for their work with pipelined 2D-Scan
of owners. For instance, if you meant to add a TV logo, based architecture running at maximum frequency of
you would think of an image as watermark more than a 100 MHz with PSNR equal to 31dB. However the extrac-
random number. tion process was not explained and no attack had been
Image to image communication algorithm is speci- ­presented to measure robustness21.
fied by6 in which the aim was to achieve distortion free 3
Proposed Spread Spectrum (SS) watermarking
communication while the original data still unaltered. scheme with two major features, first is dual watermark-
However, this attempt differs from the formers in that ing duties, one for integrity(robust watermark) and other
there is less care on carrier image quality (host image qual- for authentication purpose (fragile watermark). This had
ity) which is utilized as a high frequency carrier signal in done by embedding more than one watermark in each
digital communication, hence no quality measurements image block. Second feature was employing the watermark
values were given. for QoS measurements to check channels conditions. For
A novel six level pipelined watermarking architecture frequency transformation, Walsh transform had been
proposed by2 in spatial domain were using Reversible used while maximum frequency obtained was 80 MHz
Contrast Mapping (RCM). In this attempt, the least sig- with number of 730 Configurable Logic Blocks (CLB) for
nificant bit (LSB) is used for embedding the data after XCS40 and XCS40L chips while a gray scale image of (8*8)
applying simple integer transform on two adjacent pixels. was used to embed 4 bits watermark signal. PSNR for the
Here, watermark pixels can be reconstructed even if the watermarked image is 41.08. For authentication purpose,
LSB is lost, by one exception when the two pairs are in watermark has been tested and the alteration reflected
odd values. Two architectures are used here one for (8*8) on the watermark blocks. However, no attacks are listed
block size image and other for (32*32) with watermark for robustness evaluation. By using 4 bits as watermark
of (8*4) or (32*16) pixels. PSNR values were 29.46 dB that means there are only 16 possible watermarks may be
and 30.41 dB for Lena and Boat images, respectively. A ­created as identical case.
total of 9881 slices were needed for (32*32) bit architec- Xingfu Wang et.al22 proposed Discrete Cosine
ture with operating frequency up to 98.6 MHz, no attacks Transform (DCT) scheme to embed the random series
­measurements are shown in this work. watermark into host image, with bit error rate (BER) of
As a general notice of time domain watermarking 5%, in this work but no quality measurement is listed for
attempts, simplicity and low overhead design can be watermarked image. However, artifact which is one of the
noticed as well as real time achievement. However, we major issue in DCT transformations were noticeable on
may see in many attempts the watermark implemented the watermarked image. The IDCT is done by transfer-
by a set of random sequences, not real images, although ring the watermarked data to another board connected
the watermark can be extracted here, there must be via RS23222.
enough bits to uniquely define ownership19. Moreover, 20
Developed reversible (Lossless) watermarking algo-
attacks in many cases are either not measured, or limited rithm, they tried to achieve maximum embedding rate
to traditional ones. As a result, many researchers meant and robustness using integer wavelet transform. The
to use frequency or dual domain for their watermarking PSNR was 31.37 for embedding capacity of 27% and up
­systems. to 4 watermark bits have been embedded per coefficient.

4 Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology
 Taha Basheer Taha, Ehkan Phaklen and Ngadiran Ruzelita

The Xilinx Spartan 3 FPGA was used with frequency up ­ easurements values are listed. Components of the
m
to 62.073 MHz. Modelsim was used for simulation, but ­security keys were the original image size, wavelet trans-
the input image was a text file previously converted by form filters and the channel in which watermark is
Matlab programming environment which means the embedded, however, these factors could be easily deter-
system needs extra software support to be implemented. mined. No results for FPGA was shown26. DCT based
This, as a result, may negatively affects real time imple- approach has developed by Khoshki et.al. where auto
mentation requirement20. code generation is utilized with the help of Matlab and
An adaptive watermark system has been proposed Altera DSP builder. The code downloaded to DE2-115
by Chen Feifei et.al23 in 2012, the system is adaptive Cyclone IV FPGA27. Nallathambi, B. and Karthigaikumar,
according to image format whether it is BMP or JPEG. P. (2014) fuse Iris images and fingerprint to create the
In case of JPEG, the image is decoded into BMP before secret key, they utilized and compared DWT, DCT and
applying watermarking algorithm. Host data is a col- LSB which is spatial approach28. The comparison is done
ored image of BMP or JPEG format with PSNR between according to different attacks. However, Attacks’ factors
(66.9152 - 77.7584) while the watermark is a binary are not mentioned, for example, different noise ratio gives
image. No objective quality measurements are shown different PSNR and also the ratio of JPEG compression
for extracted watermark but subjectively, watermark was yields different results while differentiating between one
not identical but could be recognizes even after cropping algorithm and another3.
attack23. Koushik and Mahanta et.al4 (2014) proposed Message
DWT based watermarking technique for colour and State Indicator (MSI) number based image watermarking
gray scale images had been used by24 in 2012 and (256*256) architecture which relied on Walsh transform. The water-
of watermark image was used. This paper mentioned the mark can be extracted by MSI number which is generated
reason beyond using the robust watermark to overcome in embedding process. Walsh transform has been applied
issues like the high power consumption and large area for both host and watermark images. Results shows the
requirements, however for most researcher’s opinion that watermark is successfully extracted but no numeric qual-
is not the core reason to use robust watermark, since it is ity measurements are listed. In this work, watermark
used for copyright protection and confirm ownership9. No cannot be recovered in case of MSI damage4.
attacks are listed and the results show the watermark had As general scope of Transform-based image water-
been distorted after extraction. High frequency achieved marking on FPGA devices, high image quality values
by hitting frequency of 344MHz. Images is converted to could be achieved, hence more robustness. More bits are
bits using Simulink before it processed by FPGA. That’s used to construct watermark data which it is mostly an
again, affects the real time requirement and it may be a image represents owner’s logo. However, designs are more
reason that help to obtain a high frequency since not all complicated and in many cases not all the watermarking
steps are achieved using FPGA24. processes is done by hardware while researchers in some
DWT also used by A.D. Darji et.al25 when they pre- cases rely on software to apply image conversion or even
sented a new pipelined, blind FPGA-based watermark frequency transformation. Some researcher attempted to
approach using quantization method. A gray scale image mix the advantages of time and frequency domains so
is used as host image and a binary image is used as water- they used dual mode watermarking architectures.
mark. For the parallel architecture, a maximum frequency
obtained is 29.017 MHz while in pipelined approach iii.  Dual Domain Watermarking Schemes
is 97.507 MHz. PSNR for the watermarked image was Amit M Joshi et.al10 (2011) used both bit slicing scheme
44.408 dB while a range of (30-45) dB after attacks was (Spatial) and DWT (Frequency) to develop their water-
given. They compared this range with21 which is a video- marking algorithm. The approach applied lifting wavelet
image approach. It is important to mention that not all transform (to reduce the computations of wavelet con-
the steps for the algorithm were done by FPGA hardware, volutions) on (64*64) host image and applied bit space
but DWT, IDWT and the extraction processes had been slicing on the low frequency band (LL) in DWT. A binary
performed using Matlab25. watermark which is the output of LFSR is embedded on
S. Kiran et.al26 used the mid frequencies of Haar LSB plane. It required number of slices 1078 and max fre-
DWT domain to embed watermark bits. No ­ quality quency obtained on XC4VLX25-10FF676 FPGA Virtex

Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology 5
 Analysis of Current FPGA-Based Image Watermarking Attempts

was 146.651 MHz. PSNR values obtained were 61, 60,

and 59 for different images. As robustness measurement,
some non-geometric (traditional) attacks are presented10.
This work is compared with17, but this comparison has
some limitations since the later uses real images as water-
mark rather than LFSR sequence. These images need extra
operations as they used as conversion to one dimension
vector or any other processing steps. The second mixed
attempt established by Sudip Ghosh et.al17 in 2009 by fus-
ing the fast Walsh Hamdard transform to their attempt in
spatial18 to propose a dual mode watermarking algorithm Figure 3.  Peak Signal to Noise ratio for different
with Spread Spectrum concept. The xc4vlx200-11ff1513 watermarking attempts.
from Virtex 4 FPGA series had been used. Maximum fre-
quency obtained is 90.131 MHz and number of utilized
slices were 570 and 831 for spatial and transform domain,
respectively. Table 1 summaries current literature attempts
for FPGA image watermarking14.

6. Analysis
From the variety of measurements found in literature,
two common and most mentioned factors are listed in
Figures 3 and 4, PSNR and Frequency results respectively.

Table 1.  Current FPGA-Based image watermarking

attempts. Figure 4.  Maximum frequency obtained for different
Number Paper [ref.] Domain Year watermarking attempts.
1 16 TIME 2004
An important notice must be taken while examine those
2 5 TIME 2007
values is that, attempts are differs from one approach
3 17 TIME 2009 to another, that is, for example7 has high PSNR but the
4 18 TIME 2012 watermark had some distortion even before applying
5 6 TIME 2013 attacks, 24had high PSNR but only 4 bits are embedded
6 2 TIME 2014 as watermark, and in paper7 Images are converted to bits
7 21 FREQ 2009 using software before FPGA processing and that reduces
8 3 FREQ 2009
the system complexity and increase performance. For that
reason, the listed data may be taken as reference to observe
9 20 FREQ 2010
currently achieved values for frequency and PSNR, not
10 23 FREQ 2012
for comparison establishment. The same aspect is applied
11 24 FREQ 2012 on number of slices, for example, in6, DWT and IDWT
12 25 FREQ 2013 (which are the most resource consuming operations), are
13 26 FREQ 2014 not included in FPGA design process.
14 27 FREQ 2014 Furthermore, different measurements of resources are
15 28 FREQ 2014 given for different FPGA devices or families, for examples,
in Spartan devices term of Control Logic Blocks (CLB)
16 4 FREQ 2014
is used, each CLB contains two slices, each slice contains
17 10 Dual Mode 2011
four 6-input LUTs and eight storage elements in 7 series
18 14 Dual Mode 2012 of Spartan 329, while in Altera FPGA, Logic Elements (LE)

6 Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology
 Taha Basheer Taha, Ehkan Phaklen and Ngadiran Ruzelita

is mentioned instead of CLBs, LE in Cyclone III device comprehensive research aspects and results which clarify
family has a four-input LUTs, programmable registers readers’ doubts. Some of the features which are absent in
and set of connectors30. many FPGA-based watermarking schemes are listed as
Another aspect is the disparity of hardware while using follow
different types of images, coloured, grayscale, or binary
images, for example, five multipliers and six adders are • Robustness against Basic Geometric Attack (Scaling,
needed to convert RGB image to YCbCr image31. YCbCr Rotation Cropping).
conversion is a middle step for applying image conver- • Adaptable Algorithm.
sion operations such as time domain to DCT frequency • 2D-Images are utilized as watermark data.
domain, these resources are not required while using • Dual purpose scheme (Robust and Fragile)
grayscale or binary images. • Quality measurements (PSNR, MSE or any others)
While comparing with software watermarking • Pipelined design (For maximum system utilization).
attempts, there are huge difference in number of works
achieved in hardware and software. Maybe the number Table 2 lists the existence of these features in literature
is not the only factor, many FPGA-attempts lack that attempts.

Table 2.  The Existence of basic FPGA-watermarking features in literature

Watermark as Dual Quality Pipelined
Ref. No. Domain Geometric Attacks Adaptability
Image Purpose Values System
16 TIME - - - - - -
5 TIME - - - -  -
17 TIME - -  - - -
14 TIME - -  -  -
For Attacked
6 TIME - -  - -
images only
2 TIME - -  -  

21 FREQ - -  -  
3 FREQ - - - -  -
22 FREQ - - - - - 
20 FREQ - - - -  -
JPEG-BMP
23 FREQ Cropping Only  -  -
Adaptable
24 FREQ - -  - - -
Attacks as general are
25 FREQ -  -  
mentioned
26 FREQ - -  - - -
27 FREQ - -  - - -
YES (without Specifying
28 FREQ - Biometric Data -  -
attacks’ value)
Mentioned it is robust
4 FREQ against attacks without -  - - -
numbers
Within Lifting
10 Dual Mode - - - - 
based DWT
14 Dual Mode - -   - -

Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology 7
 Analysis of Current FPGA-Based Image Watermarking Attempts

7. Conclusion Journal of Emerging Trends in Computing and Information

Sciences. 2010; 2(1):37–42.
Hardware-based watermarking becomes a need with   9. Cox I, Miller M, Bloom J, Fridrich J, Kalker T. Morgan
the increase of real time image processing applica- Kaufmann: USA: Digital watermarking and steganography.
tions. Watermarking as authentication, integration and 2008.
property protection tool need more researches and 10. Joshi AM, Darji A, Mishra V. Design and implemen-
tation of real-time image watermarking. Xi’an: 2011
developments, since the limitations of resources and
IEEE International Conference on Signal Processing,
complexity of some hardware design tools bounded
Communications and Computing (ICSPCC). 2011; p. 1–5.
many researchers from presenting detailed papers which
11. Digital Water Marking Applications. Date accessed:
cover mostly all needed measurements. This paper pres- 11/1/2016: Available from: www.digitalwatermarkingal-
ents a glance of the literature attempts and analysed those liance.org/applications.asp.
works to figure out future needs of FPGA-based image 12. Huo Y, He H, Chen F. A semi-fragile image watermarking
watermarking systems. Works are classified accord- algorithm with two-stage detection. Multimedia tools and
ing to domain of embedding and extracting watermark applications. 2014; 72(1):123–49.
bits, from simple low overhead time domain to more 13. Kahng AB, Lach J, Mangione-Smith WH, Mantik S, Markov
complicated frequency domain embedding ending with IL, Potkonjak M, Wolfe G. Watermarking techniques for
dual domain schemes. Our current work aims to create a intellectual property protection. DAC’98, Proceedings of
comprehensive FPGA-based image watermarking system the 35th annual Design Automation Conference. 1998; p.
to shrink the gap between hardware and software image 776–81.
14. Ghosh S, Talapatra S, Sharma J, Chatterjee N, Rahaman
watermarking attempts.
H, Maity SP. Dual Mode VLSI Architecture for Spread
Spectrum Image Watermarking using Binary Watermark.
8. References Procedia Technology. 2012; 6:784–91.
15. Mundhada SO, Shandilya VK. Spatial and Transformation
1. Janani VSE, Ganeshkumar P. Compliant Data-centric Domain Techniques for Image Enhancement. International
Network Processing for Energy Economic Data Collection Journal of Engineering Science and Innovative Technology
in Wireless Sensor Networks. Indian Journal of Science and (IJESIT). 2012; 1(2):213–16.
Technology. 2015; 8(S9):506–12. 16. Mohanty SP, Kumara RC, Nayak S. Heidelberg: Springer-
2. Maity HK, Maity SP. FPGA implementation of reversible Berlin: FPGA Based Implementation of an Invisible-Robust
watermarking in digital images using reversible contrast Image Watermarking Encoder. Intelligent Information
mapping. Journal of Systems and Software. 2014; 96:93–104. Technology. 2004; p. 344–53.
3. Maity SP, Kundu MK, Maity S. Dual purpose FWT 17. Ghosh S, Ray P, Maity SP, Rahaman H. Spread spectrum
domain spread spectrum image watermarking in real time. image watermarking with digital design. Patiala: IACC
Computers & Electrical Engineering. 2009; 35(2):415–33. 2009, IEEE International Advance Computing Conference.
4. Mahanta K, Das DJ, Bhuyan HM, Dutta A, Gogoi M. 2009; p. 868–73.
Design and implementation of an MSI number based image 18. Ghosh S, Talapatra S, Chatterjee N, Maity SP, Rahaman H.
watermarking architecture in transform domain. Noida: FPGA based Implementation of Embedding and Decoding
2014 International Conference on Signal Processing and Architecture for Binary Watermark by Spread Spectrum
Integrated Networks (SPIN). 2014; p. 157–63. Scheme in Spatial Domain. Bonfring International Journal
5. Mohanty SP, Kougianos E, Ranganathan N. VLSI archi- of Advances in Image Processing. 2012; 2(4):1–8.
tecture and chip for combined invisible robust and fragile 19. Abdel-Hamid AT, Tahar S, Aboulhamid EM. A survey
watermarking. IET Computers & Digital Techniques. 2007; on IP watermarking techniques. Design Automation for
1(5):600–11. Embedded Systems. 2004; 9(3):211–27.
6. Maity SP, Kundu MK. Distortion free image-in-image 20. Venkateswarlu SC, Reddy PB, Raju YD. Watermarking for
communication with implementation in FPGA. AEU- JPEG2000 compression standard on FPGA. Proceedings of
International Journal of Electronics and Communications. the International Conference and Workshop on Emerging
2013; 67(5):438–47. Trends in Technology. 2010; p. 309–14.
7. Hartung F, Kutter M. Multimedia watermarking techniques. 21. Karmani S, Djemal R, Tourki R. Efficient hardware archi-
Proceedings of the IEEE. 1999; 87(7):1079–107. tecture of 2D-scan-based wavelet watermarking for
8. Gunjal BL, Manthalkar RR. An overview of transform image and video. Computer Standards & Interfaces. 2009;
domain robust digital image watermarking algorithms. 31(4):801–11.

8 Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology
 Taha Basheer Taha, Ehkan Phaklen and Ngadiran Ruzelita

22. Wang X, Qin Q, Cheng Y. Design and Implementation of Conference on Current Trends in Engineering and
Digital Image Watermark Based on FPGA. Heidelberg: Technology (ICCTET). 2013; p. 323–25.
Springer Berlin: Recent Advances in Computer Science and 27. Khoshki RM, Oweis S, Wang S, Pappas G, Ganesan S.
Information Engineering. 2012; p. 223–29. FPGA Hardware Based Implementation of an Image
23. Feifei C, Yuhui L, Bo L, Yu L. FPGA-based adaptive image Watermarking System. International Journal of Advanced
watermark embedding method and implementation. Research in Computer and Communication Engineering.
Shenzhen: IET International Conference on Information 2014; 3(5):6400–05.
Science and Control Engineering, ICISCE 2012. 2012; p. 28. Nallathambi B, Karthigaikumar P. FPGA implementa-
1–4. tion of hiding information using cryptographic key.
24. Karthigaikumar P, Baskaran K. FPGA Implementation Coimbatore: 2014 International Conference on Electronics
of High Speed Low Area DWT Based Invisible Image and Communication Systems (ICECS). 2014; p. 1–5.
Watermarking Algorithm. Procedia Engineering. 2012; 29. 7 Series FPGAs Configurable Logic Block. Date accessed
30:266–73. 17/11/2014: Available from: http://www.xilinx.com/sup-
25. Darji AD, Lad TC, Merchant SN, Chandorkar AN. port/documentation/user_guides/ug474_7Series_CLB.pdf.
Watermarking Hardware Based on Wavelet Coefficients 30. Logic Elements and Logic Array Blocks in the Cyclone III
Quantization Method. Circuits, Systems, and Signal Device Family Cyclone III Device Handbook. 2011 Dec;
Processing. 2013; 32(6):2559–79. 1:1–8.
26. Kiran S, Nadhini Sri KV, Jaya J. Design and implementation 31. Mohanty SP, Kougianos E. Real-time perceptual water-
of FPGA based invisible image watermarking encoder using marking architectures for video broadcasting. Journal of
wavelet transformation. Coimbatore: 2013 International Systems and Software. 2011; 84(5):724–38.

Vol 9 (33) | September 2016 | www.indjst.org Indian Journal of Science and Technology 9