Research On Strong Robust Watermark Algorithm Based On The Scale Space Signature Area

Digital watermarking technology,as an effective way of intellectual property protection and authentication,has been widely used.The key of watermark algorithm is to enhance watermark robustness without destroying good invisibility.Based on the analysis of various digital watermarking algorithm,aiming at the problem that the selection of feature regions in the existing algorithms is not sufficient to reflect the important information of the image,which leads to the weakening of robustness,a strong robust watermarking algorithm for feature regions in scale space is proposed.The feature points of carrier image are detected by scale space theory.The feature points are used as the center of feature region to extract the feature region which is close to the center of gravity of the carrier image and does not overlap with each other.The feature region matrix is synthesized and the watermark information is embedded by the transform domain watermarking algorithm.In the process,the frequency of the image is subdivided by discrete wavelet transform,then the low frequency subband is used as the discrete cosine transform to obtain the coefficient matrix.Finally,the diagonal matrix of the coefficient matrix after singular value decomposition is modified according to the embedding strength of the watermark.The feature region of the watermark image may be extracted and the feature region matrix is synthesized.The watermark information is extracted from the matrix by embedding the watermark inverse process.The standard grayscale image is chosen as the carrier image in the experiment,and the different watermark images are attacked with various types and different intensities.The normalized correlation between the extracted and decrypted watermark and the original watermark is above 0.98 and up to 1,and the peak signal-to-noise ratio(PSNR)of the embedded watermark is more than 45 dB.Experimental results show that the algorithm is robust and inperceptible to all kinds of attacks with different intensity.In the low frequency region near the center of gravity of the carrier image,the diagonal matrix of singular value decomposition is modified according to the embedding intensity.Not only the robustness of the watermarking algorithm can be effectively enhanced but also the invisibility is balanced by detecting the feature points in the scale space,and modifying the diagonal matrix of the singular value decomposition according to the embedding intensity.