A Robust Digital Watermarking Algorithm Based On Geometric Feature Points

The main applications of robust watermarking include resolving copyright disputes, setting user identity and data management and tracking, etc. The original watermarking technology modify the spatial domain value of digital media directly, and later developed to modify the value of transform domain. Currently digital watermarking technology has a large number of different algorithms, and they made all kinds of "robustness" of the proof and algorithms. However, most description of the "robustness" in some algorithms is nothing but a watermarking robustness against common signal processing, they can not effectively resist the general geometric attacks and complex attacks. The current digital watermarking against geometric attack is not only an important issue, but also a difficult in digital watermarking technology.This paper studies and analyzes the development status of the robust watermarking, using geometric feature points to embed digital watermarking information based on the previous work. This method can effectively deal with the problem of asynchrony after geometric attacks, in order to extract watermarking from the attack image effectively. First of all, this paper extracts candidate feature points based on Harris-Laplace algorithm, and then use their local variance to select feature points in the candidate feature points. Even after the geometric attacks, these feature points still can be extracted effectively. Then, according to the gradient direction of feature points we determine the characteristic direction, and select the feature regions through the direction of feature points and feature. Finally, by using the human visual characteristics and discrete cosine transform, we embed digital watermarking in the mid-frequency coefficients of discrete cosine transform of the feature area. The relevant experimental results show that after signal processing, general geometric attacks as well as their complex attacks, the method proposed in this paper can still detect digital watermark. So the method proposed in this paper has strong robustness after signal processing and geometric attacks.