Robust stereo images watermarking based on correlations of left and right views 


Stereoscopic video is popular because it can bring people more realistic scenes, and the application demands of the market are also driving the rapid development of stereoscopic video technology. With the continuous advancement of computer and network technologies, attackers can easily copy, modify, edit, and store multimedia content. Thus, the certain content of the high-end technical resources have certain security risks. Moreover, intellectual property rights are seriously threatened. How to effectively prevent illegal copying during the transmission of stereoscopic video content is an important issue in copyright protection.
    Digital watermarking provides a new solution to the problem of stereo images copyright protection. Stereoscopic images watermarking is generally divided into two categories. One type is to treat the stereo images as two separate monocular images, and the stereoscopic images are processed by the monocular image watermarking algorithm. Although this type of method can complete watermark embedding, the perceptual characteristics of the stereo images and the relationship between the left and right viewpoints in the stereo image are not considered. Thus a high watermark quality is not obtained. The other type needs to analyze and study the unique features of the stereo images, which can achieve preferable copyright protection of the stereo images.

    The research team of Professor Jiang Gangyi of Ningbo University proposed a robust image watermarking method based on the correlation of left and right viewpoints. Firstly, the left and right color viewpoints are separately decomposed by Tucker, which respectively generate three feature images, namely EA1、EA2、EA3 and EB1、EB2、EB3. The three feature images retain the information of the left and right viewpoints with varying degrees, and the proportion of the first feature image is particularly obvious. Secondly, in order to preserve correlation of the left and right viewpoints with the stereo images, the first feature images of the left and right viewpoints, EA1 and EB1, are combined. In order to obtain the feature image of the joint graph, P1 and P2, the tucker decomposition is performed again. P1 is the main energy image of the stereo images. The energy image contains the correlation of three-channel with the left and right color viewpoints, and preserves the correlation between the left and right viewpoints. Finally, P1 is decomposed by SVD. According to relationship of coefficients for the left singular matrix, the watermarks are embedded. The watermarked left and right viewpoints are obtained.

           
(a)The first feature; (b)The second feature; (c)The third feature

About Team
Professor Jiang Gangyi's research team is affiliated to Faculty of Information Science and Engineering, Ningbo Ningbo University, the Joint Key Laboratory of Embedded Systems, Zhejiang Province, and the Multimedia Communication Engineering Center of the Ministry of Education. The team has been working on research of 3D video processing and communication, high dynamic range imaging and processing, information hiding and so on. The team published and co-published more than 100 papers in SCI journals, such as IEEE Transactions and Signal Processing, and finished national natural science fund key projects, surface projects, international cooperation projects, etc. They completed national high-tech projects with cooperation, etc.  They authorized more than 50 invention patents in China and authorized 5 US invention patents, and successfully transferred some patents. The research team participated in the second prize of the National Science and Technology Progress Award, and won one first prize, two second prizes and three third prizes for provincial and ministerial level scientific and technological progress.

Article
Zhang Fangyan, Luo Ting, Jiang Gangyi, et al. Robust stereo images watermarking based on correlations of left and right views[J]. Opto-Electronic Engineering, 2018, 45(12): 180054. 
DOI:10.12086/oee.2018.180054