Reversible watermarking scheme for medical image based on differential evolution
Introduction
With the rapid development of telediagnosis, telesurgery as well as hospital information system, medical images have become one of the most important tools in helping physicians to determine suitable diagnostic procedures (Das and Kundu, 2012, Giakoumaki et al., 2006, Kong and Feng, 2001, Li and Kim, 2013). Medical images are also essential in evaluating patients’ recovery from their treatment (Das and Kundu, 2012, Deng et al., 2013, Giakoumaki et al., 2006, Kong and Feng, 2001). However, the sharing, handling, and processing of medical images can lead to security, confidentiality, copyright forgery, and integrity issues. Therefore, it is essential to provide security solutions for medical images to prevent any misuse or violation. To address the security issues related to medical images, Podilchuk and Delp proposed an effective and promising solution using a watermarking technique (Podilchuk & Delp, 2001). Watermarks such as patient-ID, tag, label, trademark, logo or signature is embedded into the multimedia object by changing the pixel gray level values of image without any perceptible changes on the host image (Giakoumaki et al., 2006, Kong and Feng, 2001). In fact, reversible watermarking technique or lossless method (Celik, Sharma, & Tekalp, 2006) is especially useful for medical images as it is possible to recover the original image without any distortion at the receiver side.
To allow doctors to make an accurate diagnosis using medical images, even very small distortion should be avoided in medical applications. Reversible watermarking methods not only meet the watermarking requirements of robustness, imperceptibility and capacity, they also can retrieve the host signal without distortion. Therefore, these methods have been widely applied in the literature (Alattar, 2004, An et al., 2012, Arsalan et al., 2012, Coatrieux et al., 2009, Coatrieux et al., 2006, Coatrieux et al., 2013, Deng et al., 2013, Farfoura et al., 2012, Kamran et al., 2014, Ni et al., 2006, Shi and Xiao, 2013, Shih and Wu, 2005, Tian, 2003, Zhang et al., 2013) for both copyright protection and tampering authentication in the recent decade. Due to the algebraic or geometric properties of reversible watermarking, it is especially suitable for medical images (Arsalan et al., 2012, Coatrieux et al., 2009, Coatrieux et al., 2013, Farfoura et al., 2012, Shih and Wu, 2005). Difference expansion (Alattar, 2004), sorting and prediction (Sachnev, Kim, Nam, Suresh, & Shi, 2009), histogram modification (An et al., 2012, Coatrieux et al., 2013), lossless compression (Celik et al., 2005, De Vleeschouwer et al., 2003), prediction-error histogram (Zhang et al., 2013) and hybrid methods (Kamran et al., 2014) are the most popular algorithms to realize the reversibility of watermarking algorithm. Although there are some existing work (Arsalan et al., 2012, Coatrieux et al., 2009) on the discussed topic of medical images, reversible watermarking of medical images using recursive dither modulation (RDM) still remains uninvestigated.
It is known that larger quantization steps (QSs) lead to higher robustness, but more distortion on the host images will be introduced due to the larger QSs. On the other hand, smaller QSs result in higher transparency but often lead to lower robustness (Chen & Wornell, 2001). Different medical images have different spectral components resulting in different tolerance to distortion, thus single QS would not be applicable for all host medical images. To mitigate this problem, one popular way is to insert multiple watermarks in the host image. Another way is to find the optimized solutions by trial and error. However, without any specific consideration of spectral properties of the host signals, the empirically value may lead to undesirable QS. To address this issue, a myriad of methods in the recent literature have been proposed to optimize the parameters to meet the conflicting watermarking requirements using artificial intelligence (AI) techniques (Aslantas, 2009, Aslantas et al., 2009, Fındık et al., 2011, Kumsawat et al., 2005, Liu and Tan, 2002, Run et al., 2012). This balance is achieved by formulating the watermarking algorithm as an optimization function. Consequently, many intelligent techniques such as differential evolution (DE) (Ali and Ahn, 2014, Aslantas, 2009, Lei et al., 2013b), constrained clonal selection algorithm (Aslantas et al., 2009), particle swarm optimization (PSO) (Lei et al., 2012a, Lei et al., 2013a, Run et al., 2012), and genetic algorithm (GA) (Kumsawat et al., 2005) were proposed to resolve this optimization problem effectively.
We have witnessed another trend which introduces fast intelligent watermarking scheme to reduce the computational cost in these AI based schemes (Vellasques, Sabourin, & Granger, 2013). Since the schemes in transform domain are more robust to attacks, these classes of AI techniques are usually applied in transform domain such as discrete wavelet transform (DWT) (Aslantas, 2009), integer wavelet transform (IWT) (Arsalan et al., 2012, Lee et al., 2007), lifting wavelet transform (LWT) (Lei et al., 2012b, Lei et al., 2013b), discrete cosine transform (DCT) (Aslantas et al., 2009, Lei et al., 2011), and singular value decomposition (SVD) (Run et al., 2012) rather than in spatial domain (Liu & Tan, 2002). An alternative method to improve the robustness of watermarking involves the PSO method, but it is found to be inferior to SVD-based methods. Besides, there are no security measure adopted in both DCT–SVD and DWT–SVD methods, and thus security is still a great concern for this scheme. Furthermore, the performance of existing methods (Aslantas, 2009, Kumsawat et al., 2005) with genetic algorithm is still not optimal and should be investigated further.
Striking a balance between conflicting requirements is highly dependent on the automatic selection of the important controlling parameters such as QS, threshold, scaling factor and watermarking strength. It is very common that intelligent algorithms (Aslantas, 2009, Aslantas et al., 2009, Kumsawat et al., 2005, Liu and Tan, 2002, Run et al., 2012) are utilized to obtain desirable performance by optimizing one or two parameters. However, the tradeoff among the three contradictory requirements: robustness, imperceptibility and capacity are rarely investigated. Moreover, it is reported that DE (Storn & Price, 1997) can find optimal solutions over a specified range simultaneously, and hence the best solution is achieved appropriately. In view of this, the learning abilities of DE should be exploited for QS selection, which provide two-fold benefits too. First, the selection of proper QSs is able to adaptively control watermark and achieves better imperceptibility. Second, DE is able to select QSs that provide enhanced detection (under various attacks) even without the knowledge of watermark and attack parameters.
The main goal of this paper is to design a recursive DM (RDM) based watermarking system to effectively prevent the illegal use of the medical images without affecting its visual quality. The proposed heuristic watermarking method incorporates the wavelet transform including DWT, IWT, LWT, SVD, RDM, DE and scrambling to achieve optimal performance. The singular values (SVs) of the low frequency wavelet transform coefficient are utilized to insert watermarks using optimized QSs determined by the DE heuristic algorithm. The main contributions of this work are as follows:
- (1)
Both signature and logo data as watermark are inserted by recursive dither modulation algorithm to achieve reversibility with good performance.
- (2)
Uniquely designed fitness function for DE optimization to consider all conflicting requirements rather than one or two requirements only and makes the system more adjustable. Therefore, the balance of robustness, capacity and imperceptibility is achieved by the designed parameters appropriately.
- (3)
Hybrid SVD and transform domain watermarking methods (i.e. lifting, discrete and integer wavelet) with comprehensive analysis and experiments to demonstrate the effectiveness of the proposed scheme.
- (4)
Watermarking medical image with consideration of encryption in the medical application to address the security issue of the application without encryption. A detailed security analysis for the adopted security measure is provided too.
The organization of this paper is as follows. Section 2 provides a general overview of the related work. Section 3 discusses the proposed methodology in detail. Our experiments and discussion for validating the performance of our proposed method are provided in Section 4. Finally, we conclude our paper in Section 5.
Section snippets
Related work
Digital medical images in hospital information system as well as picture archiving and communication systems that have been widely transmitted over internet can be illegally modified or duplicated (Arsalan et al., 2012, Coatrieux et al., 2009, Dandapat and Chutatape, 2004, Das and Kundu, 2012, Kong and Feng, 2001). Watermarking has been commonly applied to prevent illegal manipulation and access to the medical content without the permission of owner. The most popular methods in the field of
Differential evolution
The following steps illustrates the DE process in detail (Storn & Price, 1997):
Initialization: Define the fitness function, population size, mutation and cross rate, maximum iteration number, and optimized parameters randomly within the boundary constraints. Iterate the first generation by random selection.
Mutation: Adjust the pixel values based on each target vector, a mutant vector is generated according to:where xi,G, i = 1, 2, 3, … , D is
Experiment configuration
In our experiment, simulations are carried out to evaluate the effectiveness of the proposed reversible medical image watermarking method. Commonly used popular gray-level natural image (e.g. Lena and Baboon) are selected as reference images to compare the proposed method with existing watermarking methods. Three 512 × 512 medical images including X-ray, MRI, ultrasound (US) images, two 64 × 64 watermarks (one is text data for signature, another one is logo) are used in our simulations, which are
Conclusions
In this article, a robust and reversible watermarking scheme that embeds/extracts watermarks blindly using RDM scheme and DE optimization is proposed. The proposed method inserts double watermarks into the original host images by recursively modifying the SVs of different blocks. Furthermore, DE is utilized to optimize the QSs for controlling watermark strength with the specially designed fitness function. Overall, the proposed method demonstrates a good balance of robustness, imperceptibility,
Acknowledgements
This work was supported partly by National Natural Science Foundation of China (Nos. 61101026, 61372006, 61031003 and 81270707), partly by Shenzhen Research Project (Nos. JCYJ20120613113419607, JCYJ20130329105033277 and JSE201109150013A), partly by Project on the Integration of Industry, Education and Research of Guangdong Province and Ministry of Education (No. 2012B091100495), partly by National Natural Science Foundation of China Postdoc (No. 2013M540663), and partly by National Natural
References (54)
- et al.
An optimized watermarking technique based on self-adaptive de in DWT–SVD transform domain
Signal Processing
(2014) - et al.
Intelligent reversible watermarking in integer wavelet domain for medical images
Journal of Systems and Software
(2012) An optimal robust digital image watermarking based on SVD using differential evolution algorithm
Optics Communications
(2009)- et al.
Improving the performance of DCT-based fragile watermarking using intelligent optimization algorithms
Optics Communications
(2009) - et al.
A secure high capacity full-gray-scale-level multi-image information hiding and secret image authentication scheme via Tchebichef moments
Signal Processing: Image Communication
(2013) - et al.
A blind reversible method for watermarking relational databases based on a time-stamping protocol
Expert Systems with Applications
(2012) - et al.
A color image watermarking scheme based on artificial immune recognition system
Expert Systems with Applications
(2011) - et al.
Intelligent reversible watermarking and authentication: Hiding depth map information for 3D cameras
Information Sciences
(2012) - et al.
Robust and secure watermarking scheme for breath sound
Journal of Systems and Software
(2013) - et al.
Blind and robust audio watermarking scheme based on SVD–DCT
Signal Processing
(2011)
Optical 3D watermark based digital image watermarking for telemedicine
Optics and Lasers in Engineering
An improved SVD-based watermarking technique for copyright protection
Expert Systems with Applications
A reversible watermarking authentication scheme for wireless sensor networks
Information Sciences
Robust watermarking and compression for medical images based on genetic algorithms
Information Sciences
Fast intelligent watermarking of heterogeneous image streams through mixture modeling of PSO populations
Applied Soft Computing Journal
An intelligent watermarking method based on particle swarm optimization
Expert Systems with Applications
A novel digital watermark algorithm based on chaotic maps
Physics Letters A
Reversible watermarking algorithm of grid map based on prediction-error histogram
Journal of Software
Reversible watermark using the difference expansion of a generalized integer transform
IEEE Transactions on Image Processing
Optimal spread spectrum watermark embedding via a multistep feasibility formulation
IEEE Transactions on Image Processing
Robust reversible watermarking via clustering and enhanced pixel-wise masking
IEEE Transactions on Image Processing
Lossless watermarking for image authentication: A new framework and an implementation
IEEE Transactions on Image Processing
Lossless generalized-LSB data embedding
IEEE Transactions on Image Processing
Quantization index modulation: A class of provably good methods for digital watermarking and information embedding
IEEE Transactions on Information Theory
Reversible watermarking for knowledge digest embedding and reliability control in medical images
IEEE Transactions on Information Technology in Biomedicine
Reversible watermarking based on invariant image classification and dynamic histogram shifting
IEEE Transactions on Information Forensics and Security
Cited by (124)
An oversampling method based on differential evolution and natural neighbors
2023, Applied Soft ComputingInsights into security and privacy issues in smart healthcare systems based on medical images
2023, Journal of Information Security and ApplicationsAn efficient fragile watermarking scheme for tamper localization in satellite images
2023, Computers and Electrical EngineeringHigh-capacity reversible data hiding in encrypted images based on pixel-value-ordering and histogram shifting
2023, Expert Systems with ApplicationsPrediction error distribution with dynamic asymmetry for reversible data hiding
2021, Expert Systems with ApplicationsAdvances in medical image watermarking: a state of the art review
2024, Multimedia Tools and Applications