Removal Of Copyright Marking Based On Digital Image Inpainting

In recent years, with the rapid development of information technology and multimedia technology, the threat to digital image security is becoming more and more serious. As an important application of the digital watermarking technology, the image copyright marking is used to protect the image in a more positive way in order to solve the problem of the ownership of the digital images. But the current copyright marking in images on the internet can be easily damaged, and the robustness is still needed to be improved. Based on the contradictive requirements, this thesis explores the kind of image copyright marking which has high robustness and can’t be easily damaged by using the digital image inpainting technology to remove the image copyright marking.Firstly, this paper briefly introduces the research status of digital watermarking technology and the development of image inpainting technology, and applies the digital image inpainting technology to the removal of copyright marking. Then, by combining the different features of the image copyright marking, the copyright marking is classified; besides, the evaluation criterion and strategy of removal are given. Finally, for the different types of copyright marking, the classification removal method is designed in detail.To extract the small area copyright marking contained in images, we adopt the threshold segmentation combining with morphological operation method. After the automatic annotation, the copyright marking area is removed by the diffusion-based image inpainting algorithm. According to the different features and complexity, the large area copyright marking contained in images is divided into shape rule image copyright marking, clear outline image copyright marking and complex image copyright marking, and extracted by the manual selection, the edge detection combining with morphological operations and the saliency detection combining with threshold limit method respectively. After the copyright marking area is automatically annotated, we use the exemplar-based image inpainting algorithm to remove it.Under the condition of the color difference between copyright marking and background is larger or copyright marking located in smoother regions, the method we designed works better in extracting and removing the copyright marking area. On the contrary, the results are not good.In order to demonstrate the method for removing the image copyright marking designed in this paper more conveniently, we design a visual presentation system for the removal of copyright marking in the images based on the MATLAB platform, and give the related instructions about the system and the results after the image copyright marking removed.