Study On Robust Digital Watermarking Algorithm Based On Contourlet Transform

Since the 21st century,with the rapid rise of the Internet information era,the Internet-triggered multimedia digital information has been stolen and tampered with everywhere.At present,most digital watermarks are not robust against geometric attacks such as rotation,scaling,translation,and complex attacks such as mosaic,sharpening,and clipping.Therefore,the search for efficient and reliable algorithms to improve the watermark image's ability to resist geometric attacks and combined attacks has caused an upsurge among industry experts and scholars and has become one of the hot topics discussed by everyone.The multi-scale and multi-resolution feature of Contourlet Transform(CT)greatly improves the imperceptibility and robustness of the embedded information of digital watermarking in the transform domain.However,the second-generation watermarking scheme based on image features greatly improves the The Zernike moment makes the watermark less prone to distortion during the detection process,and thus has better robustness.This thesis combines Contourlet Transform with image features and regularized pseudo Zernike moments to successfully embed and extract watermark information,and effectively solve the problem of digital watermarking technology's weak robustness against geometric attacks.The main research work is as follows:(1)Aiming at the weak robustness of watermarked images against geometric attacks,a robust non-blind image watermarking algorithm based on CT and Singular Value Decomposition(SVD)under Blob-Harris feature regions is studied.First,the extracted low-frequency image is divided into Blob-Harris feature regions,and then the optimal feature region is selected for zero padding around the surrounding area and then normalized.The low-frequency watermark image is extracted by wavelet transform and SVD is performed to obtain the watermark information to be embedded.Finally,It is repeatedly embedded into the regular quadrilateral inscribed in the circle of each normalized feature area,and the watermark extraction effect is significant.(2)Aiming at the problem that the correct detection rate of watermarks against geometric attacks is not high,a robust digital blind watermarking algorithm based on Non Subsampled Contourlet Transform(NSCT)and pseudo Zernike moments under characteristic regions is studied.First,the extracted low-frequency image is divided into Blob-Harris feature regions,and then the optimal feature region is selected and zeros are filled around it.A stable non-overlapping square feature region is obtained as the watermark embedding region,and finally the watermark embedding region is calculated The Zernike moment and the corresponding amplitude are regularized,and the watermark is embedded in accordance with the principle of quantization and modulation.The watermark detection rate and NC_maxare relatively high.(3)Aiming at the problem that watermarking is not robust to complex attacks and combined attacks,a robust digital blind watermarking scheme combining NSCT under compressed sensing and pseudo Zernike moments is studied.First,extract the low frequency component of the original image through NSCT,obtain the sparse basis of the low frequency component through one-dimensional discrete wavelet transform,construct the measurement matrix of the sparse basis,and perform compressed sensing processing on it to obtain the new low frequency component,and then calculate its Zernike moment.The watermark information is embedded by means of jitter quantization modulating the amplitude of the regularized pseudo Zernike moment,and the NC value and bit error rate of the watermark are more ideal.