The Research On Adaptive And Resisting Geometric Attacks Image Watermarking Algorithm

In recent years, with the rapid development of network and multimedia technology, the issue to protect the copyright of multimedia becomes more and more important. As an important means of information security, digital watermarking makes up disadvantages of the traditional encryption and digital signatures, and pays a decisive role in the field of information security. Since three are contradictions between invisibility and robustness of watermarking system, this paper proposes an adaptive image watermarking algorithm to solve the problem. In addition to this, a watermarking algorithm resisting geometric attacks based on statistical characteristics is proposed, since most of them are not robust to geometric attacks.1. An adaptive odd-even quantization watermarking algorithm is proposed. A meaningful binary image watermark is processed by Arnold scrambling; JND (just noticeable difference) is composed of sensitivity, luminance masking and contrast masking based on Watson visual model in DCT domain; using JND, JPEG quantization step and a global constant, an adaptive quantization step is generated; finally, watermark is embedded in the mid-frequency coefficients of DCT domain in an odd-even way, and the coefficients are less than contrast masking. Because watermark is embedded in an odd-even way, there is no need to use original image to exact watermark. The security of algorithm is strengthened by Arnold scrambling. Experimental results show the algorithm has large capacity, good robustness in common signal processing, cropping and scaling attacks to a certain extent. The adaptive odd-even watermarking algorithm has a stronger robustness in high JPEG compression and Gaussian noise than common quantization watermarking algorithm.2. An invariant image watermarking algorithm based on statistical characteristics is proposed. In a fixed range of luminance of the image, pixels which are before and after a specified gray level are randomly selected. And these pixels are modified in order to satisfy a certain relationship to embed watermark. By judging the relationship of population of three gray levels in the same range of luminance, watermark is exacted. Experimental results show the algorithm has a strong robustness in rotation, scaling, translation attacks, large area cropping, combination geometric attacks and common signal processing. The disadvantage of the algorithm is fragile to JPEG compression and has little capacity.