Robust And Blind Watermarking Algorithms Based On Contourlet Transform And Singular Value Decomposition

The rapid growth of digital products has greatly stimulated the demand on the protection of private digital contents.Digital watermarking has emerged as a potential solution for copyright protection,which is achieved by embedding a watermark in a multimedia object,such as image,audio,or video.By analyzing the existing watermarking scheme in the transform domain,two blind digital image watermarking algorithms based on contourlte transform are designed.The main research work is as follows:A blind watermarking scheme based on contourlet transforms(CT),discrete cosine transform(DCT),singular value decomposition(SVD)and speed up robust features(SURF)algorithm is designed to resist geometrical distortion.In the watermarking embedding process,the host image is decomposed by contourlet transform and its low frequency sub-band is extracted for watermark insertion.Then the DCT is employed to analyze the low frequency sub-band and all of middle DCT coefficients are extracted.To improve the robustness of the proposed watermarking scheme,all of these DCT coefficients are tested with several common attacks and eight DCT coefficients with better stability are chosen for constructing two carrier matrices with size 2×2.Finally,the watermark sequence is embedded by modifying the largest singular values of two carrier matrices.Before the watermark extraction,a synchronized machine based on speed up robust feature is developed to evaluate the geometrical distortion parameters and correct the attacked image.Besides,the embedded watermark can be recovered without the help of original host image and original watermark.Experimental results demonstrates the presented watermarking algorithm perform well in terms of imperceptibility and can robust resist the common attacks including geometric distortion.A robust and blind watermarking algorithm based on Canny edge detection,contourlet transform and singular value decomposition are presented.The original image is divided into flat region and edge region with Canny edge detection.And it can be found that the invisibility of watermark can be improved efficiently when the watermark is embedded into flat region.For the experiments,the cover image is deal with one level contourlet transform at first.And the low frequency is divided into 8×8 non-overlapping blocks.Then,the singular value decomposition is applied on each block to generate the singular value matrix S and two matrices V and U.The watermark is embedding by modifying the coefficients in the matrix U.Besides,an extra threshold selected with least squares curve-fitting is employed in the watermark embedded stage to guarantee the robustness and invisibility.The presented watermarking scheme is tested with various attacks,and experimental results demonstrate the blind watermarking algorithm has high invisibility and robustness.