International Journal of Computer Science Engineering

and Information Technology Research (IJCSEITR)

ISSN(P): 2249-6831; ISSN(E): 2249-7943
Vol. 6, Issue 1, Feb 2016, 29-34
TJPRC Pvt. Ltd.

AN EFFECTIVE WATERMARKING SCHEME FOR 3D MEDICAL IMAGES

MANISH MADHAV TRIPATHI1 & S P TRIPATHI2
1
2

Research scholar TMU, Moradabad, Uttar Pradesh, India

Research Supervisor TMU, Moradabad, Uttar Pradesh, India

ABSTRACT
In this paper we have proposed method based on DFT and DWT transform in 3 dimensional domain. The
method can be use to authenticate and secure the 3d medical image from different types of noise attacks, compression
and geometrical attacks like median filter, translation, rotation and shear attacks. The method is useful for heftiness in
opposition to violation of different filters and pruning. Additionally, in this watermarking finding, there is no loss of
data at all i.e. lossless method. The whole process comprises ingraining and extraction dividing into small strips.
KEYWORDS: Lapel Watermarks; Heftiness; 3d Capacity Data; 3D-DWT; 3D-DFT

Received: Dec 26, 2015; Accepted: Dec 30, 2015; Published: Jan 05, 2016; Paper Id.: IJCSEITRFEB20164

In current time multimedia data, medical data, video, data, picture data are very expeditiously increasing
day by day.Moreover, in the medical field, hospitals and medical practices are going to be networked through the
internet and share medical data all around the planet. So there may be some intrusion on these moving medical
data for illegal benefits. Medical data must be protected from these illegal attacks otherwise it can harm the patient

Original Article

INTRODUCTION

as well as the medical field in terms of finance and health. Watermarking pursuites have proved to be very useful
in this regard. They can be used for authentication purposes, security purposes as per the need and sensitivity of
the data. These pursuites can be broadly divided into 3 parts, Robust watermarking, Fragile watermarking and
Semi-fragile watermarking. Additionally, in case of medical field, thereis some sensitive information which
cannot be tweaked in any ways as it will result in wreck of the image. Consequently, in this case, the image will be
divided into two parts Region of non interest (RONI) and Region of Interest (ROI). Based on this impression,
medical images can be divided into the following groups:

In the first group, watermarking in done in the RONI part as specified earlier. Several methods in this
group belong to the reversible and lossless methodalso.

In this group, those methods come which needs only authentication of the image. It can be transitioned in
both of the directions.

Academic watermarking.
These 3 methods can be exploited to provoke a healthy watermark method which is further applied with

DWT and DFT.

WEBLET TRANSFORM (IN 3 DIMENSIONS)

Following Figurerepresents the Mallet Algorithm where U, V and W represents the 3 directions of the
www.tjprc.org

editor@tjprc.org

30

Manish Madhav Tripathi & S P Tripathi

image. The image will be digitized and then applying discrete weblet transform it will be divided into two parts, High
frequency and Low frequency. Again, if High frequency contains nonuniformity, it will be further divided into two parts
High frequency and Low frequency and the same thing will happen in the case of low frequency in case of nonuniformity.
This procedure will be applied to all the three directions of the 3d capacity data of medical image.

Figure 1: 3 Directional 2 Component Decomposition

FOURIER TRANSFORM (IN 3 DIMENSIONS)

It is a signal analysis theory of discrete Fourier transform (DFT). The U X V X W3d capacity datas three
dimensions Fourier transform is defined by:
F (a, b, c) =

, ,

(1)

Where, p=0, 1,, U-1;

q=0, 1,, V-1; r= 0,1,, W-1;
f

, ,

Are the Voxel 3d capacity data R at the point (x, y, z) and F (c, d, e) be in contact to the three

dimensional DFT coefficients.

METHOD BASED ON 3D WEB-LET AND 3D FOURIER

A Scheme to Retrieve the Item Vector of Medical 3D Capacity Data
We apply the Discrete Weblet Transform method to the initial 3d capacity data of the image as given in the
above section and get the tuple PPP. Now we apply Discrete Fourier Transform on this output PPP with three Fourier
coeficients.
So this method exploits 6 coeficients 3 for DWT and 3 for DFT. As shown in the following table the method
works for different types of distortions specially geometric distortions like scaling, rotation or shear. Effect of these kinds
of distortions result in lower coefficient modifications. For high frequency positive sign has been assigned while negative
for the low frequency. Following Figurerepresents the result of different types of attacks like Gaussian Noise,

Impact Factor (JCC): 7.2165

NAAS Rating: 3.63

An Effective Watermarking Scheme for 3d Medical Images

31

Compression, Cropping, Filtering etc.

Figure 2: Effect on Frequency for Different Attacks

The ingraining and excerptin of collective watermarkswill be done are as below:
The

Watermark

will

be

provoked

to

the

group

of

binary

ersatz

contortorder.

There

arecollectivewatermarksingraining for
Bk (j). B={bk(j)|b(j)=0,1; 1jl, 1kn},

(2)

where l and n are the length and number of collective watermarks. The selected original 3d capacity data with
12812827 size. It is described as
(3)
where

!,",#

designates the choral values of 3d capacity data R at the point(i,j,k).

INGRAINING COLLECTIVE WATERMARKS

Step1:
Input item needs to process through three dimensional discrete wavelet transform (3D-DWT) & three
dimensional discrete Fourier transform (3D-DFT). First of all we convert the image into 3 dimensional matrix to get the
three coefficients to process these pursuits. After getting these 3 coefficients we apply DWT.Then,three dimension
Fourier transform (3D-DFT) of the whole pursuit sub-band ma (i,j,k), PPP1 is calculated and thethree dimensions
Fourier transform (3D-DFT) coefficient matrix. The output after applying DFT is another 3 coefficient.Moreover, letthe
frequency order O (j). It can be retrieved from low frequency to high frequency. The item vector R={r(j)|r(j)=0,1; 1jJ}
consist of the order of the low frequency three dimension discrete wavelet transform (3D-DWT)-three dimension discrete
Fourier transform coefficients, where it can be value J tune the robustness and the ability of the embedded watermark.It is
procedure describe as:

www.tjprc.org

editor@tjprc.org

32

Manish Madhav Tripathi & S P Tripathi

(4)
Step2: In this step, the collectivewatermarks Bk(j) and 3d capacity data R (j) will be employed to provoke the
logical order, vark(j);
(5)
where R(j) designates the item vector of 3d capacity data, Bk (j) designates the collective watermark to be
embedded, and is the exclusive-OR operator. The k-th var, vark (j) is correlates to k-th watermarking, Bk (j).
This coefficient will be stored for the purpose of extraction of the watermark when needed. It can be computed
k

var (j) for the logical order moreover, It can be regarded secret var and registered to be the third part preserve the
ownership of the original image.
Step 3: Extraction of 3d Capacity Data from the Collective Watermarks
Let the medical data is f(I,j,k) sent to the web after processing through Discrete Weblet Transform and Discrete
Fourier Transform.Let the variable by which it can be extracted at the receiver end is R(j). Then 3D data can be extracted
in the reverse manner as indicated in step 1. For all the slices the value is R={r (j) |r (j) =0, 1; 1jJ}. There are 6
coefficients generated, 3 for DWT and 3 for DFT
The Extraction process can be represented as follows:

(6)

Here i,j,k are the dimension of the volume.

Step 4: Collective Watermarks Bk(j) Extraction
Bk(j)=vark(j) Ex- Or R(j)

(7)

In the above equation vark(m) is the input variable stored at the time of embedding as indicated in step 1 while
Exk(m) is the extracted watermark. K in the superscript represent the kth slice of the volume data.All of the Bk(j) obtained
from the above equation will be put together one by one and we get the extracted volume data.

RESULT AND DISCUSSIONS

In our observation, it is used in absolute binary ersatz morph order100 groups. Every order is using 64 bits.In
thisobservation of four stages group are arbitrarily selected group as 20th group, 40th group, 60 group and 80 group from
100 groups asthe embedded collective watermarks.It is 3d capacity data size of 12812827. There are original 3d
capacity data represented byF (I, j, k). where 1I, j128, 1k27. Corresponding three dimension discrete wavelet
transform and discrete Fourier transform coefficient matrix is FF(I, j, k) considering the capacity embedded. For low
Impact Factor (JCC): 7.2165

NAAS Rating: 3.63

An Effective Watermarking Scheme for 3d Medical Images

33

frequency we are considering the vector R(j).

There is order to measure the quantities; similarly the NC is use in this paper defined a
(8)
Where B designate the embedded watermarking and B designates excerpting watermarking.
The amount of noise and the corresponding noises can be represented by the following figure.
In the Figure 3 below, we can see that for all of the noise types the curve of Noise parameter is linear, which
indicates that there is no change in the content information of the core medical data. Likewise PSNR value is also
approximately linear which indicates that there is no signal loss or gain at the output end of the image. Effect of different
types of attacks is explained below:

In case of Gaussian noise our method has given very good result. All the watermarks in the image were
unchanged upto Gaussian noise of 5%.

Compression upto 4% did no change in the image watermark.

For median filter attack 10 times the quality degradation was recorded only 3% down.

Figure 3: Effect of Different Types of Noise Attack

CONCLUSIONS
In our paper we have proposed method for 3 dimensional medical image data. We have applies discrete weblet
transform and discrete fourier transform. These two transform checks in all the three dimensions of the medical data and
possibility of embedding without ant distortion in the core information of the data. DWT is used because of different slices
of the data when processed it act like non stationary data. Since DWT work well for non stationary model therefore it is
applied first after applying DWT we apply DFT to strengthen the method. We embed all the information obtained from the
coefficients of these two transforms in each possible slice. At the receiver end where extraction is done all the above
process would be applied in reverse order to get the original image and watermark information. The output PSNR value

www.tjprc.org

editor@tjprc.org

34

Manish Madhav Tripathi & S P Tripathi

indicates that there is no variation in the output signal and this proves the validity of the method.
REFERENCES
1.

Manish MadhavaTripathi, S P Tripathi, Strict Attestation ofMedical Image Watermarking, International journal
ofEngineering Research & Technology (IJERT) Vol. 2 Issue 4, April 2013.

2.

Manish MadhavaTripathi, S P tripathi, A Review of MedicalImage Watermarking Schemes, InternationalJournal

ofEngineering Research & Technology (IJERT),Vol. 1 Issue 10,December- 2012.

3.

Manish MadhavaTripathi, S P tripathi, Improved Watermarkingand Recovery using Modulo DCT, TMU conference,
TMU,Moradabad, 2014.

4.

Du y., and Zhang T., A reversible and fragile watermarkingalgorithm based on DCT, 2009 International conference
onArtificial Intelligence,2, pp.301-304.IEEE,2009.

5.

Manish MadhavaTripathi, S P Tripathi, A Block based ReversibleMedical Image Watermarking,International Journal of

ComputerScience and Information Technologies, Vol.4(2) 2013,306-311.

6.

Abhishek Ranjan Pandey, Manish MadhavaTripathi, MedicalImage watermarking for Mobile Smart phones,
InternationalJournal for Innovations in Engineering, Science and Management,3d capacity 2, Issue 11, November 2014.

7.

Mahesh Prasad Tiwari, Manish MadhavaTripathi, PerformanceAnalysis of Digital Watermarking Methods for Medical
Imagesbased on PSNR and MSE, International Journal for Innovations inEngineering, Science and Management, 3d
capacity 2, Issue12,December 2014.

8.

A. Wakatani, Digital watermarking for ROI medical images by using compressed signature image, in Proc. 35th Hawaii
International Conference on System Sciences, 2002, pp.2043-2048.

9.

9.R S Yadav, Md R beg, M Mtripathi, Image Encryption Techniques: A critical comparison, International Journal of
Computer Science Engineering and Information Technology Research,2013,Vol 3, Issue 1, pp 67-74.

10. 10.M M Tripathi, Mod Haroon, Minsa Zafar, Mansi Jain, Maxillofacial Surgery Using X-Ray Based Face Recognition by
Elastic Bunch Graph Matching, Springer Berlin Heidelberg, Book: Contemporary Computing, pp183-193.

Impact Factor (JCC): 7.2165

NAAS Rating: 3.63