DOCX

Invisible_Digital_Watermarking

By Lynn Wallace,2014-05-29 00:43
11 views 0
Invisible_Digital_Watermarking

    ECE 6550 Project

    Invisible Digital Watermarking

    Based on DCT

    Chen, Shengfeng

    Jia, Zhuokang

    Instructor: Dr. Ikhlas Abdel-Qader, P.E.

    thJune 27 2010

    Western Michigan University

    THIS PROJECT IS INVISIBLE DIGITAL WATERMARKING DESIGN AND SIMULATION, USING MATLAB

    ECE6550 Invisible Digital Watermarking Based on DCT 1

    1. Introduction: ............................................................................................................................. 2

    2. Background ............................................................................................................................... 2

    3. Methods .................................................................................................................................... 3

    3.1 DCT transform ............................................................................................................... 3

    3.2 Watermarking Preprocessing ........................................................................................ 4

    3.3 Characteristics of Digital Image Watermarking ............................................................. 5

    3.4 Flow Diagram of Invisible Digital Watermarking ........................................................... 5

    3.5 Watermark embedding ................................................................................................. 6

    3.6 Blind Extraction of watermark ...................................................................................... 7

    3.7 Assessment index .......................................................................................................... 7

    4. Analysis and simulation of algorithm performance ................................................................ 8

    4.1 Test of performance after attacking. ........................................................................... 10

    4.1.1 JPEG compression ........................................................................................... 10

    4.1.2 Image Cropping ............................................................................................... 12

    4.1.3 Salt & Pepper Noise ......................................................................................... 14

    4.1.4 Gaussian Noise ................................................................................................ 14

    4.1.5 Gaussian low-pass Filter .................................................................................. 15

    4.2 Comparison between DC and low frequency AC component. .................................... 16 5. Conclusion .............................................................................................................................. 17

    Reference ........................................................................................................................................ 18

    Appendix ......................................................................................................................................... 19

    Matlab Code: ................................................................................................................................... 19

    1

    ECE6550 Invisible Digital Watermarking Based on DCT 2

    1. Introduction:

    In recent years, digital watermarking techniques have been extensively exploited and regarded as a potentially effective solution against illegal reproduction or theft of multimedia contents. Watermarking is the process that embeds information which could be data, tag or label into a multimedia object such that watermark can be detected or extracted later to make an affirmation about the object. An important classification is to divide watermarking techniques into visible and invisible according to the visibility of watermark data in embedded contents. Generally, visible watermarking schemes are used to protect digital images or videos that have to be released for certain purposes, such as contents used in distant learning web sites or digital library, while illegal copying is prohibited. On the other hand, invisible watermarking is suitable for most forms of digital contents. Users cannot perceptually recognize the difference between invisibly watermarked contents and original ones unless watermark extraction procedures are used. This method conceals both the content of the message (cryptography) and the presence of the message (steganography). An invisible watermark is very difficult to remove. Thereby, this technology could greatly strengthen the enforcement of copyright law on the Internet (Huang, C. and Wu, J.). In this project, we will introduce invisible digital watermarking and its background. We will describe the DCT transform which is an important method applied in the digital watermarking. Then, we will focus on discussing invisible digital watermarking, such as algorithm of digital image watermarking, watermark preprocessing, image embedding, and image extracting. Ultimately, according to the results of implementation, the properties of each algorithm will be analyzed.

2. Background

    Digital watermarking can be used to protect the intellectual property for multimedia data. Digital watermarking has two main advantages: firstly, the authority on a particular file can be traced; secondly, there is no need to use additional space

    2

    ECE6550 Invisible Digital Watermarking Based on DCT 3

    to maintain this information. A digital watermark is an invisible secret message that is embedded directly in a multimedia file. According to applications, watermarking systems can be classified in fragile and robust. In a fragile watermarking scheme, the watermark is designed to be fragile so as to detect and localize modifications made to the image. Robust watermarking schemes are used for copyright protection. The watermark is designed to be robust against attacks in order to protect ownership of the image. In addition, watermarking systems may be classified in non-blind detection or blind detection which decide whether the detector uses the original image to extract the watermark or not (C. Maria, G. Huiping, M. Luigi and J. Sushil). In our project, we introduce a blind image watermarking system, since only the embedded image and the secret key are required in the detection phase, and the original image is not required.

3. Methods

    3.1 DCT transform

    The discrete cosine transform (DCT) helps separate the image into parts (or spectral sub-bands) of differing importance (with respect to the image's visual quality). The DCT is similar to the discrete Fourier transform: it transforms a signal or image from the spatial domain to the frequency domain, but using only real numbers. There are multiple advantages to using the DCT even better than the Fast Fourier Transform. It has higher compression rate and less error rate. Its basis vectors are comprised of entirely real-valued components which greatly enhance efficiency. Thereby, DCT is important to numerous applications in science and engineering, from lossy compression of audio and images.

    The general equation for a 2D DCT is defined by the following equation:

    ??????:??,????:??:???????????????????????????(????(???? ??????????

The inverse transform is defined as:

    3

    ECE6550 Invisible Digital Watermarking Based on DCT 4

    ???????????,????:??????????????????????????????(??(?? ??????????

    ???????, ??;And C (u), C (v) = ????????????,??????

    Moreover, there are many advantages of the 2D DCT Transform when N is regarded as 8 or 16. After DCT transform, most of the energy in the image is compressed in the low frequency values in the frequent domain. There are two methods to DCT transform image f (m, n). One is to regard f (m, n) as 2-dimension

    matrix and do DCT transform to f (m, n) directly. Then, embed watermark in the

    image. The other method is conform to the JPEG compression standard, which divides the original image into 8?8 size blocks and do DCT transform to every single block respectively. Then, embed watermark in the image. In this project, we choose the latter methods.

    3.2 Watermarking Preprocessing

    Arnold transformation is a method to preprocess the watermark. It can disorder the image matrix and make the image illegible. However, its excellent periodicity can return the scrambled image to the original one by doing Arnold transformation of one period. One time Arnold transform is shown below:

    ((??????????????;?; ? (?)?(?)??,??????????? ))???

    ? x and y denote the coordinate of pixels of the original watermark, (,~; )?

    denote the coordinate of pixels of the transformed watermark, N denotes the size of original image.

    watermarkArnold Transfer watermark

    To see the example shown above is the original watermark. The right one is the

    4

    ECE6550 Invisible Digital Watermarking Based on DCT 5

    watermark image after 5 times Arnold transformation

    3.3 Characteristics of Digital Image Watermarking

    1) Visible and Invisible

     Visible watermarking schemes are used to protect digital images or videos, such as contents used in distant learning web sites or digital library, while illegal copying is prohibited. However, invisible watermarking is suitable for most forms of digital contents. Users cannot recognize the difference between invisibly watermarked contents and original image unless watermark extraction procedures are used 2) Fragile and Robustness

     The watermark is designed to be fragile so as to detect and localize modifications made to the image. However, robust watermarking schemes are used for copyright protection. The watermark is designed to be robust against attacks in order to identify ownership of the image

    3) Security

     A strong algorithm applied to the watermark embedding should be able to withstand different attacks and retain the watermark without destroying. 3.4 Flow Diagram of Invisible Digital Watermarking

    Watermarking Watermarking Watermarking

    Embedding Extraction Preprocessing

    Main frame of invisible digital watermarking

     Arnold Secret Watermarking

    Key Information Transform

    Algorithm of Embedded Embedding Image Watermark

     DCT Original Transformed

     Image Image

    Theory of Watermark Embedding

    5

    ECE6550 Invisible Digital Watermarking Based on DCT 6

Secret

    Key Extracting Algorithm of

     Watermarking Detecting

    Information Watermark Embedded Image

    Theory of Watermark Extraction

3.5 Watermark embedding

    In this project, we will choose 512?512 gray scale image and 64?64 watermark

    as an example.

    1) Load the original image and watermark W

    2) Turn watermark into binary

    3) Preprocess watermark by using Arnold transformation with 5 times

    ???4) Divide original image into 8?8 blocks ?, then do DCT transformation to

    every single block from left to right, up to down in sequence so as to obtain

    ??????????????=DCT(?)=????????????

    5) Embed watermark into DCT transformed DC component.

    ??Assume ?=?? ???? ?. R is quantized value which is used to adjust ??

    depth of embedding. R would reduce the stability of embedded watermark

    if its value is too small. However, too big value of R will decrease the quality

    of original image as well. Thus, we will choose 32 as the value of R in order

    to implement a good result. The code below is the method we embed

    information of watermark.

     %When watermark M(m,n) is 1:

    if M(m,n)==1

     if Wk<=R/4

     block_dct1(1,1)=block_dct1(1,1)-Wk-R/4;

     elseif P>R/4&&P<=2*R/4

     block_dct1(1,1)=block_dct1(1,1)-Wk+3*R/4;

     else

     block_dct1(1,1)=block_dct1(1,1);

    6

    ECE6550 Invisible Digital Watermarking Based on DCT 7

     end

     end

    %When watermarking information M(m,n) is 0:

     if outM(m,n)==0

     if Wk>=3*R/4

     block_dct1(1,1)=block_dct1(1,1)- Wk+5*R/4;

     elseif Wk<3*R/4&& Wk>=2*R/4

     block_dct1(1,1)=block_dct1(1,1)- Wk+R/4;

     else

     block_dct1(1,1)=block_dct1(1,1);

     end

     end

    6) Inverse DCT transformation to the block that has been embedded with

    watermarking information

    7) Repeat step 5 and 6 until all of the watermarking information have been

    added to the all of the blocks.

    3.6 Blind Extraction of watermark

    ?Define ? is the image embedded watermarking information, n is the times of Arnold transformation.

    1) Divide the embedded image into 8?8 blocks, then do the DCT

    transformation.

    2) Make sure we know exactly the value of R, then detect DC coefficient and

    extract watermarking information:

    if mod(block_dct1(1,1),R)>=R/2

     J(m,n)=1;

     else

     J(m,n)=0;

     end

    3) Repeat step 2 until all of watermarks have been extracted from the blocks

    4) Do 5 times inverse Arnold transformation, then extraction is completed.

    3.7 Assessment index

    In practice, Peak signal-to-noise (PSNR) is a very important index to assess quality

    of embedded image.

    ??PSNR = -10lg ?

    7

    ECE6550 Invisible Digital Watermarking Based on DCT 8

    And

    ?????????????D =????(?!? , !?( ??????????

    ( is the value of pixel of the original image, denotes the value of pixel of output ?

    image, N denotes the numbers of pixels in the output image. [0, M-1] is the coverage of image pixels.

    Correlation coefficient denotes similarity between extracted watermark and original watermark, which determine robustness is whether good or not.

    ??????((??????????,?

    ????????(?????

    ?W denotes original watermark, denotes extracted watermark which had been

    ??attacked. ( and ( denote the value of pixel of and W respectively. ??

    4. Analysis and simulation of algorithm performance This project uses original image with size of ??????? and watermark with size of

    ????? as shown.

    original image

    watermark

     Original image watermark

    Adopting R=32, 5 times Arnold transformation, embedding watermark in DC

    8

    ECE6550 Invisible Digital Watermarking Based on DCT 9

    coefficient. The embedded image is shown below.

    embeded image

    Embedded image

    In order to evaluate the performance of the watermark algorithm, there are at least three aspects that need to be tested.

    (1) Hidden Property.

    It is a contradiction between the amount of information and hidden extent in

    digital image watermarking. As the increase of the amount of watermarking

    information, the quality of the image must be declined.

    (2) Robustness.

    The process of testing robustness is usually done by attacking an embedded

    image. This test evaluates the ability that if watermark can resist various

    attacking. These types of attacking include compression, cropping, filtering and

    adding noise.

    (3) Security.

    Security is an index to evaluate the complexity and runtime of extracting

    watermark.

    9

Report this document

For any questions or suggestions please email
cust-service@docsford.com