Digital Image Watermarking Algorithm In Region Of Interest

Science and Technology is a "double-edged sword". Its development has brought convenience to people's lives, and it also provided convenient conditions for lawbreaking activities. In the information age, whether digital images worthy of trust or not has aroused greatest concern. The electronic information that stored in digital form spread rapidly in the internet makes piracy, tampering and illegal acts of destruction happen easily. In this case, digital watermarking emerged. Nowadays, with the development of the second-generation digital watermarking technology, attacking on image watermarking system is becoming more complex. So it requires that watermarking must have stronger performance against geometric attacking. Generally, the common image watermarking attacking is signal processing, such as median filtering, sharpening, noise adding, and JPEG compression. And general geometric attacking can be described as rotation, translation, scaling, bending at any angles and shear attacking. Common signal processing weakens the strength of the watermark, but geometric attacking makes synchronization error between the original and extracted watermark. Previous image watermarking scheme shows robustness toward common signal processing, but become powerless for the general geometric attacking. In view of this awkward situation, the watermarking technology resistance to geometric attacking has developed rapidly in recent years. These programs generally include Template-based approach, Image invariant features, and affine invariance of moments for the original image and so on.In this dissertation, we present an algorithm against geometric attacks on the region of interest in the image. In order to present this algorithm, I researched lots of existing algorithms that against geometric attack. This algorithm can protect the area to be protected and improve the efficiency of the watermarking algorithm.Firstly, this dissertation proposes digital watermarking on the texture area of image, embeds the watermark signal into the texture area of image, and then proposes a robust digital image watermarking scheme based on the mechanism of visual attention. By the models of visual attention it gets the regions of interest (ROI) of the cover image, and then hides the watermarks in the texture block of the ROI. It gets the saliency map of the cover image by the models of visual attention, and then constitutes the area to embed watermark from the saliency map and the texture block, and finally hides the mark in the certain area by adjusting the coefficients of Tchebichef moment to protect the ROI. It realizes automatically segment the ROI of the image.Secondly, this dissertation presents a watermarking scheme against geometric attacking based on scale invariant feature transform. In the first place, we use SIFT operator to detect the parameters of geometric attacks suffered by the image. In the second place, we restore the damaged image according to the parameters. Lastly, we extract the watermark from the restored image.