Research On Remote Sensing Image Map's Copyright Protection Based On Feature Points

With the rapid development of the network technology, digital map products for the storage and application are even more extensive. As a new digital map product, remote sensing image map is widely used in the field of military mapping and geographic information navigation. How to protect copyright of remote sensing image map products has become a major research focus. Faced with this problem, people have proposed digital watermarking technology to protect the copyright of the products. Digital watermark technology loads the watermark such as text, binary image or a sequence of the number into the digital products by a particular way in order to protect the copyright. There are several problems in the remote sensing image map's watermarking algorithms:It is difficult to determine the location for embedding watermark, past approaches choose some positions in the pixels or the frequency domain of the map to embed the watermark directly. These approaches could have good effects in a few special cases of attacks, but are not versatile; The perturbation method of watermark doesn't have good performance. The perturbation method has a direct impact on the amount of the change in remote sensing image map's data; The characteristic of the remote sensing image map is not considered in the watermarking algorithms. Most watermarking algorithms are just designed using the data of the image.To solve the above problems, combined with the remote sensing image map's characteristics, this dissertation presents a blind watermarking algorithm for copyright protection. The algorithm uses the stability of harris feature points and the characteristics of the wavelet multi-level to get the feature points on the field at all level of low frequency wavelet coefficient. The feature points are unified in the same dimension and classified into several clusters to get the representative points. Then a representative point of the cluster is selected to locate the reference point. The representative points got by this method are more relocatable. And this method enhances the robustness of the algorithm. At last the scrambled binary image information is embedded into the wavelet coefficient by the proposed perturbation method. The proposed perturbation method ensures the invisibility in maximum extent also improves the resistance to attacks. The invariance of the attribute data is considered in the extraction process. The proposed algorithm uses two methods to get the watermark. First method uses the geographical coordinates of the attribute data as a secret key to get watermark. While another way gets the watermark uses the embedding method as a reference. The proposed extraction method not only ensures the efficiency of the algorithm but also protects the integrity of the extracted watermark.The experimental results show that the proposed algorithm has good performance of the invisibility and is effective against the geometric attack, filtering attack, noise attack, jpeg compression attack.