Image watermark embedding method and extraction method, and related methods

Abstract

The present invention discloses an image watermark embedding method, an image watermark extraction method, a digital media copyright protection method and a digital media copyright tracing method. The image watermark embedding method comprises: converting a watermark into a watermark binary sequence as required; carrying out a key-based spread spectrum operation on the watermark binary sequence, so as to determine a spread spectrum watermark vector; carrying out adaptive segmentation on a to-be-watermarked image; based on the adaptive segmentation, carrying out multistage discrete cosine transform on the to-be-watermarked image, so as to determine a first transform coefficient matrix; based on the first transform coefficient matrix, constructing a feature vector, and embedding the spread spectrum watermark vector into the feature vector; and carrying out coefficient reset, inverse discrete cosine transform of the coefficient matrix, and the like on the feature vector embedded with the spread spectrum watermark vector, so as to determine the watermarked image. According to the methods disclosed by the embodiments of the present invention, invisibility, robustness and security of the watermark are improved, and risk that the watermark is recognized and erased maliciously is weakened.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of digital media copyright protection, in particular to an image watermark embedding method, an image watermark extraction method, a digital media copyright protection method, and a digital media copyright tracking method. Background technique [0002] In recent years, with the rapid development of Internet technology and digital media technology, there are a large number of digital media content resources (such as text, images, audio and video, etc.) Media content is copied, modified, disseminated and used. Therefore, the copyright protection of digital media content has attracted more and more attention from academic and commercial circles. In the process of solving this problem, adding watermark to digital media content is a widely used technical means. Digital watermarking technology is to embed the identifier representing its copyright information into the digital media conten...

AI Technical Summary

Problems solved by technology

The main feature of the spatial domain algorithm is to realize the watermark embedding by directly modifying the pixel domain coefficient of the image or by adjusting the statistics of the image pixel value. Basically, it does not have the ability to resist geometric attacks; the transform domain algorithm is to transform the image pixel domain coefficients into a certain frequency domain (mainly including discrete Fourier transform domain, discrete cosine transform domain, discrete wavelet transform domain and singular value decomposition, etc.) , and then modify the frequency domain coefficients according to a specific rule, so as to achieve the purpose of embedding the watermark. Some of the algorithms also increase the modulation of the visual masking model in the watermark embedding process according to the characteristics of the human visual system, so that the invisibility of the watermark is further improved. ; At the same time, there are watermarking algorithms based on spread spectrum communication and watermarking algorithms based on dirty paper grid coding; the above types of algorithms can improve the watermark's ability to resist image processing attacks to varying degrees, but for rotation, scaling, flipping, etc. Geometric attack, the above algorithm basically has no resistance

[0004] With the further development of digital image watermarking technology and further expansion of application fields, the current research focus on image watermarking algorithms is mainly focused on improving the algorithm's resistance to geometric attacks. Algorithms that have been proposed can be roughly divided into four categories: one is based on geometric The algorithm of invariant domain, this type of algorithm has a strong ability to resist geometric attacks, but a large number of interpolation operations and data truncation operations will be introduced in the process of geometric transformation and inverse transformation, which will have a great impact on the visual quality of the image. The main reason that hinders the further development of this type of algorithm; the second is based on the image normalization algorithm, which can also resist geometric attacks such as rotation, scaling, and flipping to a certain extent, but theory and experiments show that this algorithm is only for square images. The normalization effect is better, and the normalization of rectangular images is difficult to ensure the consistency before and after the attack, which greatly limits the application range of this type of algorithm. The third is the algorithm based on template embedding and matching. This kind of algorithm greatly improves the ability of the algorithm to resist geometric attacks due to the use of templates, but a significant problem is that the embedded template points are easy to be analyzed and analyzed. It is maliciously erased, thus losing the synchronization mechanism, which is the main reason why this type of algorithm cannot be directly applied to the actual digital copyright protection system; the fourth is the algorithm based on image feature points and feature areas, which can better It is resistant to attacks such as rotation and scaling, but due to the particularity of this type of algorithm, most of the algorithms belong to the space domain algorithm, and the ability to resist image processing attacks is relatively weak. The dependence on the nature of the algorithm is relatively large, which also makes the anti-attack performance of this type of algorithm greatly restricted.

Images