Research On Some Key Issues For Digital Image Watermarking

With the rapid development of Internet and multimedia technology,people can get multimedia resources more conveniently.At the same time,the illegal copying,forgery and dissemination of these resources are becoming easier,which undoubtedly poses a serious threat to the legitimate rights and interests of the copyright owners.As an active authentication technology,digital image watermarking has become an important means of copyright protection for image resources.However,there are still two difficult problems in the field of digital image watermarking:one is to enhance the robustness to resist geometric attacks,the other is to achieve good imperceptibility in the premise of resisting geometric attacks.In order to solve the two problems effectively,this paper proposes some effective image watermarking algorithms.In addition,this paper proposes a new set of invariant orthogonal moments,namely polar harmonic Fourier moments(PHFM),which can be effectively used in image watermarking.The main contents can be summarized as follows:(1)In order to reduce the time complexity and improve the computation accuracy of traditional method for radial harmonic Fourier moments(RHFM),a fast and precise method by using fast Fourier transform is proposed and based on the which,a novel image watermarking algorithm which is robust to geometric attacks is presented.The RHFM of the original image is firstly computed by using the proposed method and the robust RHFMs which are suitable for watermark embedding are selected.Then the watermark is embedded by adaptively modifying the magnitudes of RHFM based on quantization.In the decoder,the watermark can be extracted from the magnitudes of RHFM directly without using the original image.Experimental results show the proposed algorithm provides an excellent watermark invisibility and can be resilient to geometric attacks and common image processing attacks effectively.(2)A robust color image watermarking algorithm based on fuzzy least squares support vector machine(FLS-SVM)and Bessel K form(BKF)distribution is presented,which is a recently developed geometric correction algorithm.Firstly,the quaternion discrete Fourier transform(QDFT)of the maximum central region of the original color image is performed.Then the watermark is embedded into the magnitudes of low-frequency information of QDFT.In watermark decoding process,the synchronous correction based on FLS-SVM model is used.When training the FLS-SVM model,the quaternion wavelet transform(QWT)of the grayscale images that correspond to the color training images is firstly performed,and BKF distribution is then used to fit the empirical histogram of coefficients of the QWT,and the shape parameters and scale parameters of BKF distribution are finally used to construct image feature vector.Experimental results show that the proposed algorithm is not only invisible,but also has outstanding robustness against common image processing attacks and geometric attacks.(3)A new set of invariant orthogonal moments,namely polar harmonic Fourier moments(PHFM)is proposed,which can be effectively used as image features in the image analysis.PHFM does not suffer numerical instability problem,whereas Zernike moments(ZM),pseudo Zernike moments(PZM)and radial harmonic Fourier moments(RHFM)suffer it.For the same degree,the RBF of PHFM has more zeros than that of ZM,PZM and polar sine transform(PST)and these zeros are more evenly distributed.This means the image description ability of PHFM is superior to that of ZM,PZM and PST.Moreover,due to the complex exponential function and trigonometric function can be transformed into each other,PHFM can be considered as a generalization of polar complex exponential transform(PCET),polar cosine transform(PCT)and PST.Theoretical analysis and experimental results show that PHFM has more excellent performance compared with ZM,PZM,RHFM,PCET,PCT and PST in terms of image reconstruction and rotation invariant object recognition under the conditions of noisy and various attacks.(4)A new blind multiplicative watermarking algorithm is proposed.Due to the the strong robustness against various attacks,polar harmonic Fourier moments(PHFM)magnitudes are employed as the watermark carrier.The PHFM magnitudes are highly non-Gaussian and a proper distribution to model the statistics of the PHFM magnitudes is a heavy-tailed probability density function(PDF).It is proved that Weibull distribution can suitably fit the distribution of PHFM magnitudes and based on this,a statistics-based watermark detection by using the Weibull as a prior for the PHFM magnitudes is proposed.In watermark embedding,watermark information is embedded into the PHFM magnitudes of highest entropy blocks using multiplicative manner so as to achieve better robustness and imperceptibility.In watermark detection,a blind statistical watermark detector using the maximum likelihood(ML)decision rule based on the Weibull distribution is developed.Compared with Bessel K form(BKF),Cauchy and generalized Gaussian(GG)-based decoders,the Weibull-based decoder demonstrates a stronger robustness.In addition,the proposed watermarking algorithm provides an excellent watermark imperceptibility and can effectively resist geometric attacks and common image processing attacks.(5)Three zero-watermarking algorithms based on orthogonal moment are proposed,which are robust to geometric attacks.Firstly,quaternion exponent moments(QEM)for color images and ternary polar harmonic Fourier moments(TPHFM)for stereo images are derived based on orthogonal moment and hypercomplex number theory.Then,PCET,QEM and TPHFM are used to construct zero-watermark image for grayscale image,color image and stereo image respectively.Experimental results show that the proposed zero-watermarking algorithms have excellent imperceptibility,and can effectively resist geometric attacks and common image processing attacks.