Research On Digital Image Transform Domain Robust Watermarking Algorithm

In recent decades, with the advancement of multimedia technology and the network, the digital multimedia, such as the digital image, the video, music, etc, has made great prosperity in our everyday life. It is much easier than ever before to download the multimedia from the internet or present the product adveristment with digital multimedia, which makes our daily life and work much convenient and comfortable. In general, the nonshared multimedia can only be obtained by the authorized users. However, the digital multimedia could be replicated, modified and used by the unauthorized users for their business activity, which brings enormous potential loss to the originator. Therefore, more and more attention is paied to the intellectual property rights of the digital media and digital authentication becomes more and more prominent. As an effective way of protecting the intellectual property rights of digital media, digital watermarking has received much attention in the field of information security. The digital watermarking technique hiding the secret information (the owner or source) in the digital media to obtain the cover image for users. In practice, although the cover image may be demaged by intentional or unintentional uses, the watermark information can still be extracted to validate the ownership of the digital media.In this thesis, we studied the robust watermarking algorithms for digital images with transform domain method. The fractional Krawtchouk transform, the discrete trinion Fourier transform, the fractional Bessel-Fourier moment, and the singular value decomposition have been investigated. According to the properties of each transformation, their transform domain image watermarking algorithm was designed, followed by the imperceptibility and robustness of the watermarking scheme. The specific research contents and contributions of this thesis are briefly outlined as follows:(1) A robust image watermarking was designed with the fractional Krawtchouk transform.Based on the eigenvalue decomposition method, we generalized the traditional Krawtchouk transform to its fractional order formation. Theoretically, we determined the eigenvalues and their corresponding multiplicity of the Krawtchouk transform matrix.Moreover, a set of orthonormal eigenvectors of the transform matrix was obtained. After that,a robust watermark embedding and extraction method was designed with the proposed fractional Krawtchouk transform. Since the traditional Krawtchouk transform is a special case of the fractional Krawtchouk transform, and different transform domain can be obtained by adjusting the fractional orders, the proposed fractional transform has more potential application.Finally, the proposed watermarking imperceptibility and robustness to common signal processing attacks was validated in the experiments. A comparative analysis has also been made to the robustness effect of different fractional orders. Experimental results showed that the proposed watermarking scheme has better watermark robustness and imperceptibility than traditional Krawtchouk based scheme, when properly choosing the fractional orders in the fractional Krawthouck transform. Moreover, by comparing with the existing other transform domain watermarking algorithms (DCT, DWT, etc.), the proposed scheme has more ability to resist the most common signal processing watermark attacks.(2) Color image zero-watermarking algorithm based on singular value decomposition and visual cryptograph.First of all, a set of color image robust feature was constructed from the singular value decomposition domain of discrete wavelet transform (DWT) domain non-overlapping blocks by taking into account the fact that the low pass subband DWT coefficients are more robust to image noise, filter, and compression. To achieve the singular value decomposition of 2-D matrix,the three channels of each color block are converted to the 2-D matrix. The robust feature is generated by thresholding the first singular value of each block. Then, the robust image feature can be seen as one secret share of the watermark, and the other secret share of the watermark can be derived from the watermark and the image feature by taking the advantage of the (2, 2)visual cryptography scheme. Finally, the second secret share could be stored in certificate authority for identifying the ownership of the image. Experiment results shows that the proposed watermarking scheme has better watermark robustness than some existing congener method.(3) Color image watermarking algorithm with the discrete trinion Fourier transform.First of all, the unit trinion of the trinion Fourier transform was extended to a general form,which is then used to define the discrete trinion Fourier transform and its inverse transform.The computation the discrete trinion Fourier transform was achieved by the discrete Fourier transform of the three channels. Then the color image watermarking algorithm was designed by modifying the trinion Fourier transform coefficients of the image. Comparing with the common quaternion Fourier transform based watermarking scheme, experimental results showed that the trioion Fourier transform not only has better robustness and watermark capacity,but also avoids the energy loss problem existing in the quaternion based scheme.(4) Fractional Bessel-Fourier moment based image watermarking method resisting to geometric attacks.The conventional Bessel-Fourier moment was generalized to the fractional order Bessel-Fourier moment, and its geometric (rotation and scale) invariants have been investigated. After that, a robust color image watermarking method was designed with the invariants to resist the watermark geometric attacks and the combination attacks of geometric attacks and common signal processing attacks. When properly choosing the fractional order, experimental results showed better robustness can be achieved with the proposed method than the Bessel-Fourier moment based scheme and other watermarking scheme using existing fractional moments, such as fractional radial shifted Legendre moment, fractional orthogonal Fourier mellin moment, and generic polar complex exponent transform.