Digital Watermarking Algorithm Based On Image Local Feature

With the technological development of communication technology and multimedia technology, much multimedia digital information, such as images, video, audio, etc. is very easy to be copied and tampered by illegal users. Another problem which is the authenticity of the ownership of the multimedia digital information is extended out. To solve this problem, the digital watermarking technology is born at the right moment. The digital watermarking technology is used to represent the authenticity of the ownership of the multimedia digital information by embedding watermark information without affecting the quality of the multimedia digital information. In a subsequent experiment is determined whether watermark information exists through a special watermark extraction algorithm. There are many types of attacks, however, which try to undermine and remove the watermark information so that it cannot be properly extracted, especially, in a variety of attacks geometrical attacks are most serious destruction of the watermark information.To solve this problem, by the research of digital watermarking algorithm on geometric invariance of image local features it is proposed a digital watermarking algorithm which is named Normalized Image Watermarking based on Local Feature-point, i.e. NIWLF algorithm. First, the traditional SIFT algorithm is improved for application in digital watermarking algorithm to reduce the computational complexity and improve the operation efficiency. Using improved SIFT algorithm finds, screens and distributes local feature points in the original image,and uses local feature points to determine circular local feature areas,i.e. LFA, in certain conditions. By normalizing LFA, the watermark is embedded into the frequency coefficients in the DCT domain of LFA. Combining the SIFT algorithm which has affine invariant and using scale space and direction of the local feature points as parameters of the direction and parameters of the radius so that the circular local feature areas have scale and rotation invariance. Circular local feature areas have translational invariance, so selecting the circular as a local feature area has the same affine invariant with triangle, quadrilateral and robustness of geometric attack. Otherwise, in order to improve the balance of invisibility and robust of watermarking algorithm and take into account the impact on the image quality of the watermark information, the algorithm dynamically adjusts the watermark embedding strength based on the quantitative relationship of the embedding strength and image quality evaluation standard indicators PSNR and Watson human visual system. The experiments show that the algorithm is not only to ensure better watermark visibility, but also show strong robustness to geometric attack.