A Review of Wavelet-domain Watermarking Techniques
Mukul Salhotra, Shivam Maurya, Shashvat Jaiswal, Amit Kumar Singh Department of CS/ICT, Jaypee University of Information Technology

Abstract--As networking continues to grow exponentially, the Intellectual Property Rights (IPR) can be obtained and reproduced easily. These threats create a high demand for a content protection technique such as digital watermarking; which is one of the most efficient ways to protect the digital properties in recent years. Image watermarking techniques are frequently applied in the transform and spatial domains to achieve desired secure and robust protection. This paper provides an overview of the wavelet-based watermarking techniques available for medical images until now. In this paper the major methods have been analyzed along with their advantages & disadvantages. Keywords: Discrete Wavelet Transform, Medical image watermarking, ROI, NROI,
I. INTRODUCTION

D

igital watermarking has emerged as a research area that was originally focused on copyright protection. Also, it has been implemented in lot of domains, such as video, audio, image, and 3D graphic model. Despite the fact there are only a few medical oriented watermarking studies in the literature to date, digital watermarking will be a valuable tool for copyright protection with medical confidentiality protection, patient and examination-related information hiding, data integrity control and source identification, in Hospital Information System(HIS) and picture archiving and communication system(PACS) based on Digital Imaging and Communications in Medicine(DICOM) standard pave the way to store medical image and search for data base and give remote medical treatment [1]. Digital watermarking is the technology that embeds directly additional information by modifying imperceptible either the original data or some transformed version of them [2]. It is divided into two basic categories spatial and frequency. In recent years digital watermarking has been thoroughly researched and distinguished as a potentially effective means for protecting copyright of digital media, since it makes possible the embedding of secret information in the digital

content to identify the copyright owner. The basic idea behind digital watermarking is to add a watermark within the host data with an aim of broadcast monitoring, access control, copyright protection etc. A watermark can be a digital signal, tag or label. A host can be video, image or audio. We can classify watermarking techniques into different groups according to visibility domain. In telemedicine, it is imperative to maintain the quality of images due to their diagnostic value. This is why algorithms which generally embed watermarks into coefficient of different sub-bands with different coefficients must be improvised. They are not robust to common signal processing. In relation with this matter, the development of a new algorithm that can satisfy both imperceptibility as well as robustness is needed. Performance of watermarking schemes can be augmented by multiple methods [3].
II. ATTRIBUTES OF DIGITAL IMAGE WATERMARKING



The requirements for image watermarking can be treated as the characteristics, properties or attributes of image watermarking. Different applications demand different properties of watermarking. Requirements of image watermarking vary and result in various design issues depending on image watermarking applications and purpose. These requirements need to be taken into consideration while designing watermarking system. There are basic five requirements [4]. Fidelity: Fidelity is considered as a measure of perceptual transparency or imperceptibility of a watermark. It indicates the similarity of a watermarked and non-watermarked image. Robustness: Watermarks should not be removed unintentionally by simple image processing operations or intentionally. Hence the watermarks should be robust against variety of such attacks. Data Payload: It is the maximum amount of information that can be hidden without degrading image quality. This property describes how much data should be embedded as a watermark so that it can be successfully detected during extraction.







Security: Secret key has to be used for embedding and detection processing case security is a major concern. There are three types of keys used in watermark systems: private-key, detection-key and public-key. Computational Complexity: Computational complexity indicates the amount of time watermarking algorithm takes to encode and decode. To ensure security and validity of watermark, more computational complexity is needed [5].
III. DISCRETE WAVELET TRANSFORM



An original image can be decomposed of frequency districts of HL1, LH1, HH1. The low-frequency district information also can be decomposed into sub-level frequency district information of LL2, HL2, LH2 and HH2. By doing this the original image can be decomposed for n level wavelet transformation.
A. CHARACTERISTICS OF DWT

Discrete Wavelet transform (DWT) is a mathematical tool for hierarchically decomposing an image. It is useful for processing of non-stationary signals. The transform is based on small waves, called wavelets, of varying frequency and limited duration. Wavelet transform provides both frequency and spatial description of an image. Unlike conventional Fourier transform, temporal information is retained in this transformation process. Wavelets are created by translations and dilation of a fixed function called mother wavelet. This section analyses suitability of DWT for image watermarking and gives advantages of using DWT as against other transforms. The basic idea of discrete wavelet transform (DWT) in image process is to multi-differentiated decompose the image into sub-image of different spatial domain and independent frequency district [6] [7]. Then transform the coefficient of sub-image. After the original image has been DWT transformed, it is decomposed into 4 frequency districts which is one low-frequency district(LL)and three high-frequency districts(LH,HL,HH). If the information of lowfrequency district is DWT transformed, the sub-level frequency district information will be obtained. A twodimensional image after two level DWT decomposed can be shown as Fig.1. Where, L represents low-pass filter, H represents high-pass filter.

The wavelet transform decomposes the image into three vectors: horizontal, vertical and diagonal. Hence wavelets reflect the anisotropic properties of the Human Vision System (HVS) much more precisely when compared to with other techniques.

● ●

●

● ●

●
B.

Wavelet Transform is computationally efficient and can be implemented by using simple filter convolution. With multi-resolution analysis, image can be represented at more than one resolution level. Wavelets allow image to be described in terms of coarse overall shape and details ranging from broad to narrow. Magnitude of DWT coefficients is larger in the lowest bands (LL) at each level of decomposition and is smaller for other bands (HH, LH, HL). The larger the magnitude of wavelet coefficient, the more significant it is. Watermark detection at lower resolutions is computationally effective because at every successive resolution level, less no. of frequency bands are involved. High resolution sub bands help to easily locate edge and textures patterns in an image.

ADVANTAGES OF DWT

●

● ● ●
Figure 1 Multiwavelet sub-bands

The watermarking method supports multi resolution characteristics and HVS. Since human eyes are not sensitive to minute changes in edges and textures though they are extremely sensitive to the minute changes in the smooth parts of an image. With DWT, the edges and textures are usually subjected to the high frequency sub bands. Large frequencies in these bands indicate sharp edges in the image. It has an immense advantage over DCT is it more robust since it supports additive noise, rescaling /stretching, and half toning. It supports the current compression techniques like JPEG 2000. DFT and DCT are full frame transform. Hence, any change in the transform coefficients affects entire image except if DCT

is implemented using a block based approach. However DWT has spatial frequency locality. It means it will affect the image locally, if watermark is embedded.
C. WORKING PRINCIPLE OF DWT a) WATERMARK EMBEDDING AND EXTRACTION:

IV.

PROPERTIES OF MEDICAL IMAGE WATERMARKING

The security of medical information is essential to the privacy of the patient and to the legislative rules. This inflicts three mandatory characteristics: confidentiality, reliability and availability. Security risks of medical images can vary from random errors occurring during transmission to lost or overwritten segments in the network during exchanges in the hospital networks. It is also imperative that one must also guarantee that the header of the image file always matches that of the image data. In addition to these unintentional modifications one can envision various malicious manipulations to replace or modify parts of the image, called tampering. The usual constraints of watermarking are invisibility of the mark, capacity, secrecy to unauthorised persons, and robustness to attempts to suppress the mark. These demands also exist in the medical domain but additional constraints are added. Three main objectives are foreseen in the medical domain [9] ● Imperceptible ● Integrity Control ● Authentication
A. REGION OF INTEREST (ROI)

A number of watermarking techniques have been developed. Any watermarking algorithm follows certain common steps. For an image to be watermarked by DWT first discrete wavelet transform is applied on the original gray scale image. As figure 1 shows, secret/public key and watermark along with transformed image is processed by the embedding algorithm. After the algorithm have done embedding, inverse discrete wavelet transform (IDWT) is applied to get the watermarked image [8].

Figure 2 Watermark embedding For detection of watermark in a test image first it is transformed in DWT domain. As shown figure 2.the watermark and key along with transformed test image are a then used to find the authorization of the image.

Characteristically, a medical image is diagnosed before being archived in long-term storage, so the important part of the image has already been determined. This important part is called ROI (Region Of Interest), which must be preserved without any loss of information.

Figure 3 Watermark Detection and extraction

Figure 4 Option for selecting ROI. [26] In general, the ROI is stored as it is or compressed by a lossless algorithm and the other part is compressed by a lossy algorithm; this is done since it can achieve a

higher compression rate than that of a lossless compression algorithm [22].
V. PREVIOUS WORK DONE

A number of earlier works are available in literature format, on digital image watermarking scheme that employs dwt for image processing and watermarking of digital medical images. Here, some of the recent motivating researches are briefly described as below. C.Woo et al [10] proposed a multiple watermarking method which is suitable for privacy control and tamper detection in medical images. A new scheme was proposed in which embedding of four types of watermark was proposed to provide medical information systems with an additional level of security and physicians with an added value tool for accurate diagnosis and efficient treatment planning. These watermarks include the digital signature of the doctor for the purpose of source authentication, EPR of the patient, the ROI as a robust watermark and finally a fragile watermark which is embedded in the RONI in the wavelet domain and in the ROI in the spatial domain to provide information on whether or not the image is tampered. Ehab F. Badran, Maha A. Sharkas, and Omneya A. Attallah [11] ,the algorithm is divided into 3 steps which are; watermark generation, watermark embedding and finally watermark extraction. The watermark generation step is as follows; first the ROI after its extraction from the image is decomposed into three wavelet decomposition levels. The binary image is then divided into even and odd bits and embedded as a robust watermark into the horizontal and vertical detail coefficients of the 3rd decomposition level respectively. This binary image is embedded again as a fragile watermark in the LSB of the pixel values of ROI in the spatial domain. After the generation of the four watermarks, they are embedded into the host image. These multiple watermarks are embedded into the image using a private key depending on the ROI locations. Applying the inverse DWT will form the watermark image. The embedding step is done using [1] which is similar to the one used in [8], but the only difference is that it was used there to embed a single watermark in the spatial domain, while here it is used to embed multiple watermarks in the wavelet transform domain. Shikha Tripathi, Nishanth Ramesh, Bernito A, Neeraj K J[12], propose a new embedding approach wherein, the watermark pixels are chosen pseudo-randomly, besides selecting the locations for embedding the watermark in the mid-frequency region of the source image. This

increases the security by two-fold. The highlight of the process is that both blind and non-blind methods are incorporated into one watermarking scheme .Thus a two-level security is achieved by actually using only one watermark. This rules out any malicious attack. To further increase the security a pseudo random number generator PRNG is used at various instances in the algorithm. N. Kaewkamnerd and K.R. Rao [13] developed a wavelet based image adaptive watermarking scheme. Embedding is performed in the higher level sub-bands of wavelet transform, even though this can clearly change the image fidelity. In order to avoid perceptual degradation of image, the watermark insertion is carefully performed while using HVS. Bo Chen and Hang Shen [14] developed a new robust fragile double image watermarking algorithm using improved pixel-wise masking model and a new bit substitution based on pseudo-random sequence. The method embeds robust and fragile watermark into the insensitive part and sensitive part of wavelet coefficients making two watermarks non-interfering. A Lavanya and V Natarajan [15] proposes a watermarking algorithm which enhances medical images and data security .In this scheme least significant bit substitution is adapted to fulfil the requirements of data hiding and authentication in medical images. The scheme divides image into two parts ROI and non-ROI. The DS embedded randomly into the border of the LSB brain image using standard encryption technique. The ROI is embedded in the LSB centre of the brain image. The results demonstrate the effective embedding and retrieval of DS and ROI from the watermarked image, and the results of robustness to different attacks are found to be satisfactory. S. M. Ramesh & A. Shanmugam [16] presented an efficient image watermarking technique to protect the copyright protection of digital images with watermark embedding and watermark extraction. The results obtained from the experimentation shows that the proposed watermarking techniques provide better results with higher accuracy. Xiaonian Tang et al. [17] have proposed a simple robust watermarking algorithm which can confirm the copyright without original image. The system was that the watermarks are generated during the embedding procedure but not predetermined. The original image was transformed using Haar wavelet and flags are created with the secret key. Then the extracted flags using the secret key are compared with the original ones to confirm the copyright of digital products. Theoretical

analysis and simulation results show that the presented watermarking system has excellent robustness against various watermark attacks including slightly geometric modifications with the high quality of the watermarked image. Hamid Amiri et al. [18] have presented a robust watermarking algorithm in wavelet transform domain. Firstly, original image was decomposed into its subbands using 3D wavelet transform, then, significant coefficients with the same position in HL, LH and HH sub bands of the last level are extracted to make a triplet. This affects the robustness of algorithm and the quality of image. To specify the optimal values for constants a Multi-Objective Genetic Algorithm (MOGA) was used. Experimental results reveal high robustness of the algorithm against common image processing attacks (blurring, median filter, sharpening, noise addition and more), and print and scan distortion. Wen-Tzeng Huang et al. [19] have proposed a method of robustness and blind extraction watermark for static images. It utilizes discrete wavelet transform and applies three coding methods according to the different characteristics of band coefficients. Together, these methods embed a watermark while maintaining image fidelity. Experimental results indicate that the approach is high robust against frequency-based and time domain geometric attacks. Additionally, since this approach produces a blind watermark and thus neither the original image nor any of its related information is needed, it seems to be a very convenient and practical watermarking technique for applications. Akiyoshi Wakatani [20] proposes a simple scheme for the ROI in medical images using compressed signature images. The basic scheme is compression of the signature image by a progressive coding thus generating a bit stream in order of significance. This bit stream is then embedded into the pixels around the ROI in a spiral manner. They have also proposed a concatenation algorithm for multiple embedding of the signature image. The key points to be noted in this paper are that the signature image can be extracted with moderate quality and by dividing the contour of ROI into some parts; the signature can be generated from the clipped image including only a part of the ROI. The SNR and the quality of the recovered image has been found to be lower than usual. S. Dandapat et al [21] proposes a wavelet based data embedding schemes in the medical images. A HVS approach based on contrast sensitive function has been used to estimate the quality of the embedded medical image. The points to be noted are that embedding data into the mid and high frequency sub bands of the host

image are showing low values of DDM and high values of PSNR thus suggesting that mid & high frequency coefficients are suitable for embedding data. OHyung-Kyo Lee et al [22] proposes an embedding method which embeds the watermark with the ROI information into the adjacent area of the ROI medical image. The researchers have stated that their extracted watermark is a visually recognizable pattern. They also seem to have compared the NC, PSNR and RMS of the watermarked and non watermarked images. The attacks are successful in disrupting the image although the ROI can still be extracted. The proposed technique has successfully survived the JPEG lossy compression. Baisa L. Gunjal et al [23] have focused upon improving the DWT algorithm given by Na Li et. al [24] and increasing the security of the watermarked image while maintaining a constant PSNR value of the key improvement implemented is the Arnold transformation in which the given image is restored to its original state after passing through a series of iteration. In the embedding technique, initially the ROI was separated from the RONI, and the image was embedded in the later. The performance of algorithm has been evaluated by two metrics, Perceptual transparency and Robustness. The calculated PSNR value met the industrial standards.
CONCLUSION

There is a need of healthcare industry to transmit medical images safely and without compromising their security aspect. This paper explains the philosophies, methodologies characteristics and requirements of DWT in digital image watermarking. We support the demand for a setting up of a benchmark in relation with digital image watermarking in telemedicine [25]. The PSNR values of the various algorithms indicate that DWT algorithms have shown superior resistance to JPEG compression and visual compression. A limitation of this review is that combined attacks are not analyzed.
REFERENCES [1] C. Coatrieux, H. Maitre, B. Sankur, Y. Rolland, and R. Collorec, "Relevance of Watermarking in Medical Imaging," 2000 IEEE EMBS Conf. On Information Technology Applications in Biomedicine, pp.250-255, Nov. 2000. [2] Wan Adnan, W.A.; Hitam, S.; Abdul-Karim, S., Tamjis, M.R., “A review of image watermarking”, Research and Development, SCORED, 2003, pp. 381-384. [3] P. Fakhari, E. Vahedi and C. Lucas, Protecting Patient Privacy From Unauthorized Release of Medical Images Using a Bio-nspired

Wavelet-based Watermarking Approach. Processing, vol.21, 2011, pp.433-446.

In:

Digital

Signal

[18] S. Hamid Amiri, Mansour Jamzad, “A Robust Image Watermarking Method in Wavelet Domain Using Genetic Algorithm”, International Conference on Availability, Reliability and Security (AReS), March 2009, Fukuoka, Japan [19] Wen-Tzeng Huang, Sun-Yen Tan, Yuan-Jen Chang, Chin-Hsing Chen "A discrete wavelet transform based robust watermarking for copyright protection" Proceedings of the 12th international conference on Networking, VLSI and signal processing,Wisconsin, USA, 2010. [20] Akiyoshi Wakatani, Digital Watermarking for ROI Medical Images by Using Compressed Signature.Proceedings of the 35th Hawaii International Conference on System Sciences - 2002. [21] S. Dandapat, J. Xu, Opas Chutatape, S. M. Krishnan , Wavelet Transform Domain Data Embedding in a Medical Image,Proceedings of the 26th Annual International Conference of the IEEE EMBS ,2004 [22] H.K. Lee et aI., ROI medical image watermarking using DWT and bit-plane, Proc. Asia-Pacific Conference on Communications, 2005,512 - 515. [22] OHyung-Kyo Lee, Hee-Jung Kim, Ki-Ryong Kwon, Jong-Keuk Lee l, ROI Medical Image Watermarking Using DWT and Bit-plane, Asia-Pacific Conference on Communications, Perth, Western Australia, 2005 [23] Baisa L. Gunjal, Suresh N. Mali, ROI Based Embedded Watermarking of Medical Images for Secured Communication in Telemedicine, International Journal of Computer and Communication Engineering 2012. [24] Na Li Na Li Xiaoshi Zheng; Yanling Zhao; Huimin Wu; Shifeng Li Algorithm of Digital Image Watermarking Based on Discrete Wavelet Transform”, Electronic Commerce and Security, International Symposium ,Volume, Issue , 3-5 Aug. 2008 , [25] K. A Navas and M. Sasikumar, Survey of Medical Image Watermarking Algorithms, 4rth International Conference: Sciences of Electronic, Technologies of Information and Telecommunications March 25-29, 2007 – TUNISIA.

[4] Cox, I.J., M.L. Miller, and J.A. Bloom, Digital Watermarking. 2002, San Francisco: Morgan Kaufmann Publisher. [5] Jian Liu , Xiangjian He , “Literature Review of Wavelet Based Digital Image Watermarking Techniques” International Journal of Computer Application(0975-8887) Volume 31– No.1, October 2011. [6] GhoutiL, BouridaneA and Ibrahim MK, “Digital image watermarking using balanced multiwavelets” , IEEE Transactions on Signal Processing, 54(4), pp. 1519-1536,2006. [7] Reddy AA, Chatterji BN, “A new wavelet based logowatermarking scheme” , Conf. Pattern Recognition letters,26(7), pp. 1019-1027, 2005. [8] S, Maurya, A. Singh, M. Salhotra, Digital Image Watermarking Techniques: An Overview,National Conference on Artificial Intelligence and Agents: Theory & Applications December, BHU Varanasi,2011. [9] R. Raul et aI., Hiding scheme for medical images , Proc .17th International Conference on Electronics, Communications and Computers, 2007, 32-32. [10] C.Woo,J. Du, B.Chen,." Multiple watermark method for privacy control and tamper detection in medicalimages," Proc. APRS Workshop on Digital Image Computing (WDIC), Australia, 2005,5964. [11]Ehab F. Badran, Maha A. Sharkas, and Omneya A. Attallah,”Multiple Watermark Embedding Scheme in Wavelet-Spatial Domains Based on ROI of Medical Images “26th NATIONAL RADIO SCIENCE CONFERENCE, NSRC'2009 Future University, 5th Compound, New Cairo, Egypt, March 17-19,2009 ,Arab Academyfor Science & Technology, Alexandria. Egypt. [12]Shikha Tripathi, Nishanth Ramesh, Bernito A, Neeraj K J,”A DWT based Dual Image Watermarking Technique for Authenticity and Watermark Protection “,Signal & Image Processing : An International Journal(SIPIJ)Vol.1, No.2, December 2010 [13] N. Kaewkamnerd and K.R. Rao, "Wavelet based image adaptive watermarking scheme", IEE Electronics Letters, vol.36, pp.312-313, 17 Feb.2000. [14] Bo Chen, Hong Shen, “A New Robust-Fragile Double Image Watermarking Algorithm”, Third IEEE International Conference on Multimedia and Ubiquitous Engineering,2009 [15]A Lavanya and V Natarajan “Enhancing security of DICOM images during storage and transmission in distributed environment “, Sa¯dhana¯ Vol. 36, Part 4, August 2011, pp. 515–523.c Indian Academy of Sciences. [16]S. M. Ramesh & A. Shanmugam “An Efficient Robust Watermarking Algorithm in Filter Techniques for Embedding Digital Signature into Medical Images Using Discrete Wavelet Transform “,European Journal of Scientific Research ISSN 1450-216X Vol.60 No.1,2011, pp. 33-44. [17] Xiaonian Tang , Quan Wen and Guijun Nian “An Improved Robust Watermarking Technique in Wavelet Domain” 2010 Second International Conference on MultiMedia and Information Technology