1

CHAPTER 1 INTRODUCTION

1.1

GENERAL The protection of intellectual property has become a major problem

in the digital age. The ease of copying digital information without any loss of quality violates the conservation of mass property of traditional media, which inhibited wide global distribution in the past. On the Internet today, it is possible to duplicate digital information a million-fold and distribute it over the entire world in seconds. These issues worry creators of intellectual property to the point that they do not even consider to publish on the Internet. More information is transmitted in a digital format now than ever, and the growth in this trend cannot be estimated in the future. Digital information is susceptible to be copied at the same quality as the original. A watermark is a pattern of bits inserted into a digital image, audio or video file that identifies the file's copyright information (author, rights, etc.). The name “watermark” is derived from the faintly visible marks imprinted on the organizational stationary. During the 18th century watermarks began to be used as antiimitation measures on money and other documents. When sharing information on the internet, digital watermark approaches are of great demand. While distributing information through online, we never know if someone uses them without our knowledge. The owner should be able to hide some information in the digital file and extract information to prove his

2

ownership when the need arises. Watermarking system can be viewed as a communication system consisting of three main elements: an embedded, a communication channel and a detector. To make use of the Human Visible System (HVS), various watermarking techniques have been developed. Figure 1.1 shows a general watermarking life cycle. Tracking of reproduced copies, prevention of illegal copying and validating the digital data can be done by the watermark. Insertion of a watermark, detection of a watermark and removal of a watermark are the three main processes involved in a watermarking system. Secure Part – Transmitter

In secure Part

Secure Part – Receiver

Decoding

Embedding

Attacks Original Image

Embedding Scheme

Detection

Retrieval

Figure 1.1 General watermarking life cycle Important characteristics of the watermark are invisibility, robustness, readability and security (Ming-Shing et al 2001, Sin and sung 2001). Requirement for digital watermarks are 1) deterioration of the quality of digital content is minimized 2) watermarks are retained and detectable after the digital content is edited, compressed, or converted 3) the structure of a watermark makes it difficult to detect or overwrite (alter) the embedded information (watermark contents) 4) processing required for watermarking and detection is simple 5) watermark information embedded in digital content can be detected as required and 6) embedded watermark information cannot

3

be eliminated without diminishing the quality of the digital content that carries the watermark. 1.2

CLASSIFICATION OF WATERMARK General classification of the watermark is shown in Figure 1.2.

Watermarking is classified based on working domain, type of document, human perception and application. Watermarking techniques are divided into four categories in accordance with the type of information (document) to be watermarked (Laurence and Ahmed 1996). They are text watermarking, image watermarking, audio watermarking and video watermarking. In text watermarking, the text documents can be watermarked by patterning the interword spaces. Text watermarking is primarily of three types:

Line Shift

Coding (LSC), Word Shift Coding (WSC) and Feature Coding (FC). These methods require the original unmarked text for decoding. WATERMARKING

According To Working Domain

Spatial Domain

Frequency Domain

According To Human Perception

Invisible

Robust

Private

Public

Visible

According To Application

Source Based

Destination Based

Fragile

Invertible

Noninvertible

Quasiinvertible

Figure 1.2 General classification of watermarking

Non quasiinvertible

4

In image watermarking technique, the watermark image is applied into a host image for security. Image watermarking is nothing but the still image watermark. A continuous frame of image is called as video and the watermarking process is known as video watermarking. Digital video watermarking uses the inherent properties of digital images, with the limitations of human vision to insert invisible data into digital video to provide copyright protection. Based on the visibility of the resultant image, the digital watermarks can be divided into two different categories viz. visible watermark and invisible watermark. Visible watermark is a secondary translucent overlaid into the primary image but in an invisible digital watermarking, information is added as digital data. In the case of audio watermarking, to hide the watermark and make it inaudible, watermarking uses the time and frequency masking properties of the human ear. Echo hiding is one of the techniques which involve hiding information within the recorded sound by introducing very short echoes. Invisible digital watermark is further divided into private and public watermarking. In private watermarking or informed watermarking, the original image is required to perform the extraction process. In public watermarking or blind watermarking, the original image is not required to perform the extraction process. In the public watermarking process, watermarked images are seriously destroyed and the detection of watermarked image is very difficult. Because of this, blind watermarking technique is used for visible watermarking. Based on the ability of the watermark to resist attack, watermarks are categorized into two types. They are fragile watermark and robust watermark. Random image processing methods can readily destroy the fragile watermarks. Most of the image processing methods are robust and can be extracted from heavily attacked watermarked image without destroying the image. This makes the robust watermark to be preferred in copyright

5

protection. Invisible–robust watermark is embedded in such a way that the alterations made to the pixel value are perceptually not detected and it can be recovered only with appropriate decoding mechanism. Invisible–fragile watermark embedding process of the image would alter or destroy the watermark. Watermarking techniques are frequently used in the still camera images, medical images and satellite images where the copyright protection is required by the users. A digital image is usually represented by a two-dimensional image. Depending on the image resolution, an image may be a vector or a raster in type. Digital image usually refers to raster images and it is also called as bitmap images. Various available digital image file types are Joint Photographic Groups (JPG), Graphic Interchange Format (GIF), Tagged Image File Format (TIFF), Portable Network Graphics (PNG), and Bitmap (BMP). TIFF is a very flexible format that can be lossless or lossy. PNG is a lossless storage format; it can be used to compress the file size. The compression is exactly reversible, so the image is recovered exactly. GIF is lossless only for images with 256 colour or less. JPG works by analyzing images and discarding kinds of information. BMP is an uncompressed proprietary format. Digital Still Camera (DSC) records the image data in the form of document, specified as the standard file format. Nowadays, digital documents can be distributed via the World Wide Web (WWW) to a large number of people in a cost-efficient way. There is a strong need for security services in order to keep the distribution of digital multimedia work both profitable for the document owner and reliable for the customer. Watermarking technology plays an important role in securing the business as it allows placing an imperceptible mark in the multimedia data to identify the legitimate owner, track authorized users via fingerprinting (Dittmann 1999) or detect malicious tampering of the document (Kundur and Hatzinakos 1998).

6

Patient records are stored in hospitals in digital format (Electronic Patient Records (EPR)) for more than 20 years. Medical image has three binding security characteristics, such as confidentiality, availability and reliability. In our proposed work on confidentiality, we use public key. Availability can be proved by decoding the watermarked image by using normal procedure. Reliability will be proved by the information which cannot be modified by an unauthorized person. Reliability is of much importance as degradation of the image content will lead to serious problems such as wrong diagnosis of a patient by the doctor. Thus, watermarking is important in case of medical images for determining authenticity. Remote sensing satellite images are important sources of geographical data. Geographical data are commonly used to classify earth land cover, analyze crop conditions, assess mineral, petroleum deposits, and quantify urban growth. Contrast stretching, flipping and format conversion are the attacks that easily remove the watermark image in a satellite image. An effective watermarking technique for satellite images should have the following features: The watermark should be imperceptible to the naked eye. The watermark must be indelible, at least without visibly degrading the original image. Retrieval of the watermark should explicitly identify the owner. The watermarking technique should not distort certain specific areas in the image. Stir mark is commonly used to evaluate the robustness of an image (Evelyn et al 2009). 1.3

WATERMARKING TECHNIQUES Digital image watermarking schemes mainly fall into two broad

categories: spatial domain and frequency domain techniques. Visible watermarking mainly uses spatial domain which requires less computation and are easy to implement in software as well as hardware. A spatial domain technique slightly modifies the pixels. However, there must be tradeoffs

7

between invisibility and robustness, and it is hard to resist common image processing and noise. Some of the spatial domain modulation techniques are Least Significant Bit (LSB), Spread Spectrum Method (SSM). In LSB, the watermarks are embedded in the least significant bit of the selected pixels of an image. This method is easy to implement and it is not very robust against attacks. SSM based watermarking algorithms embed the information by linearly combining the host image with a small pseudo noise signal, which is modulated by the embedded watermark. Compared to spatial domain methods, frequency domain methods are more widely applied. In frequency domain, the characteristics of the HVS are better captured by the spectral coefficients. For example, HVS is more sensitive to low frequency coefficients and less sensitive to high frequency coefficients. Low frequency coefficients are perceptually significant, which means alterations to those components might cause severe distortion to the original image. On the other hand, high frequency coefficients are considered insignificant and hence the processing techniques, such as compression, tend to remove high frequency coefficients assertively. To obtain a balance between imperceptibility and robustness, most watermark algorithms are embedded in the midrange frequencies. Commonly used frequency domain techniques are DCT, Discrete Fourier Transform (DFT) and Discrete Wavelet Transform (DWT). DCT based watermarking techniques are robust compared to spatial domain techniques. DCT algorithms are robust against simple image processing operations like low pass filtering, brightness and contrast adjustment, blurring, etc. DCT watermarking techniques are difficult to implement and are computationally more expensive. They are also weak against geometric attacks like rotation, scaling, cropping, etc. DCT watermarking can be classified into global DCT watermarking and block

8

based DCT watermarking. DCT based watermarking are affected by two factors. The first fact is that the most important visual part of the image lies at low frequency sub-band. The second fact is that high frequency components of the images are usually removed through compression and noise attacks. DCT watermark is therefore embedded by modifying the coefficients of the middle frequency sub-band. DFT of a function gives quantitative results of the frequency content in terms of magnitude and phase. This result is more important for processing and analysis of signals and images. The DWT is currently used in a wide variety of signal processing applications, such as in audio and video compression, removal of noise in audio, and the simulation of wireless antenna distribution (Evelyn et al 2009). In wavelets, basal functions are used to represent the signal. DWT is very suitable to identify the areas in the host image, where the watermark image can be embedded. Wavelets have their energy concentrated in time and are well suited for the analysis of transient and time-varying signals. Watermarking techniques have got a number of applications. Some of the significant applications are fingerprint, prevention of unauthorized copying, image authentication, data security, digital media management, medical area and copyright protection. Copyright protection is probably the most common use of watermarks today. Copyright owner information is embedded in the image in order to prevent others from alleging ownership of the image. Copyright-related applications based on robust watermarking techniques were discussed by many researchers like Barni et al (2002), Moulin and Ivanovic (2003), Sebe and Domingo (2003), Trappe et al (2003). Medical reports play a very important role in the treatments offered to the patient. A mix up in the reports of two patients could lead to a disaster. To avoid this problem, visible watermarking technique is used to print the names of the patients on the X-ray or Magnetic Resonance Image (MRI) scan reports. Fragile or semi-fragile watermarks are usually selected for

9

watermarking process in medical, forensic and intelligence or military applications (Barreto et al 2002, Li and Yang 2003, Li 2004, Wong and Memom 2000, Xie and Arce 2001). In data encryption (embedding), techniques of digital watermarking do not follow with the same capability because listening, accessing and viewing the content cannot be prevented. For this reason, digital watermarking is not protected from hacker attacks (Yeung et al 1998). Some of the intentional attacks on watermarks are active, passive, forgery and collusion attacks (Cox et al 2000). In active attacks, the hacker removes the watermark or makes it undetectable. In passive attacks, the hacker can easily identify the presence of watermark in the original image without any damage or removal. The hacker attempts to embed a valid watermark of their own rather than removing the original watermark in forgery attacks. One piece of the media is replicated into several copies, each with a different watermark, in order to construct a copy with no watermark due to collusion attacks. 1.4

PERFORMANCE ANALYSIS Performance analysis is needed to determine the characteristics of

the watermarking technique such as imperceptible, indelible, statistically undetectable and easily decodable. Popular metrics used for evaluating imperceptibility of the watermark are Signal-to-Noise Ratio (SNR) and Peak Signal-to-Noise Ratio (PSNR), which are based on Mean Square Error (MSE) between the original and watermarked images. Image manipulation tool (stir mark) is used to measure the effectiveness of watermark embedding technique in terms of its robustness and data integrity criteria.

Pixel based visual

distortion metrics (Kutter and Petitcolas 1999) are used for performance analysis to test the image quality between the original and the watermarked images.

10

Correlation coefficient is essential for mapping and ranging purposes. Individual quality measures are not reliably associated with the strength of treatment effect in medical areas. Although the use of specific quality measures may be appropriate in specific well-defined areas of the medical field, it cannot be generalized to all clinical areas or meta-analysis (Pan et al 2004). Normalized Correlation Coordinate (NCC) computes the similarity measurement between the original watermark and the extracted watermark. Image Fidelity (IF) is a process used to deliver an image accurately, without any distortion or information loss. IF output depends upon the ability to detect the difference between images (Klimeck et al 2002). If the difference between an original image and a compressed one cannot be detected, then it is concluded that the compression is a lossless compression. SNR measures are easy to estimate the quality of a reconstructed image compared to the original image. Peak signal of the reconstructed watermark image is measured by PSNR. PSNR values are measured in decibels. Typical PSNR values range between 20dB and 40dB. The actual value is not meaningful, but the comparison between two values for different reconstructed images gives a measure of quality. MSE gives the results of degradation, which was introduced at the pixel level. The higher MSE shows more degradation. Accuracy Rate (AR) is used to measure the difference between the original watermark and the recovered one. AR is computed as follows: AR= CP/ NP. Where NP is the number of pixels in the original watermark and CP is the number of correct pixels. 1.5

PROBLEM FORMULATION This thesis aims at developing an efficient hardware architecture for

the implementation of visible watermarking technique in both spatial domain

11

and frequency domain and also aims at the performance analysis of the algorithm used for invisible watermarking technique using MATLAB 7.6. The vector based visible digital image watermarking algorithm using 1D-DCT is tested and implemented with reduced computational complexity and resource utility involving the scaling embedding computational complexity. With this implementation, the speed and throughput are increased. In biomedical applications, small distortions in the host image make more problems while diagnosing the diseases. On focusing biomedical applications, a new block based visible image watermarking algorithm is developed. In block based a fast 1D-DCT is used to reduce the resource utilization. In addition, a new mathematical model is introduced to find the values of scaling and embedding factors. Quality image can be obtained by means of combining various watermarking techniques. A new watermarking system is designed to combine the spatial and the frequency domain techniques. For the above proposed works, we developed a novel high performance VLSI architecture implemented on FPGA, simulated in Xilinx ISE 10.1 and tested in Xilinx Virtex V XC5V1X330 technology. In order to achieve high throughput and speed, the architectures are designed with the implementation of pipelining and parallelism techniques. The hardware architecture designed is applied for visible watermarking technique only. In order to touch the other category of watermarking, namely the invisible watermarking, a movement based watermarking algorithm is developed. The performance analysis of this algorithm is obtained using MATLAB 7.6.

12

Initially, the performance analysis of both visible and invisible watermarking scheme were computed using the software MATLAB 7.6 and the evaluation of the works based on synthesis were done using the Xilinx tool ISE 10.1. Finally, the throughput for visible watermarking is compared with that of the existing hardware implementation. 1.6

THESIS ORGANIZATION Chapter 2, “Literature review”, presents a detailed literature review

of the digital watermarking, the existing watermarking algorithms and the spatial and frequency domain watermarking techniques. It also presents the reviews related to the hardware implementations. Chapter 3, “Design and VLSI implementation of vector based visible image watermark using 1D-DCT”, describes the architecture design for vector based digital image watermarking. For this, the algorithm is designed to aim at reducing the computational complexity involving the embedding and scaling factors prominently used in any visible watermarking technique. Chapter 4, “High performance VLSI architecture for block based visible image watermarking”, explains VLSI architecture design and implementation of block based visible image watermarking algorithm and its performance analysis. The fast 1D-DCT for watermarking process is introduced to facilitate the hardware implementation. Chapter 5, “Design and implementation of hybrid VLSI architecture for visible spatial and frequency domain watermarking”, describes VLSI architecture design and implementation of visible spatial and frequency domain watermarking algorithms and their performance analysis. Based on choice of watermark, the process is done either as a pixel by pixel

13

operation under spatial domain or as a vector form of operation under frequency domain. Chapter 6, “Performance analysis for geometrical attacks on digital image watermarking”, describes performance analysis for geometrical attacks on digital invisible image watermarking. Here, the irreversible watermarking approach robust to affine transform attacks is used. In this approach, watermark embedding and extraction are carried out with respect to an image normalized to meet a set of predefined moment criteria. Chapter 7, “Conclusion”, summarizes the contribution of this thesis by the implementation of the three proposed approaches in the hardware, the algorithm proposed for invisible watermarking and its performance evaluation. Suggestions for future work are also included.