Research On DIBR 3D Video Copyright Protection Algorithm

With the development of digital multimedia technology and the improvement of network communication levels,the threshold of 3D video transmission and production is further reduced.The 3D video synthesized by depth image-based rendering(DIBR)has attracted extensive attention because of its small volume,easy popularization,and strong compatibility.Then there are copyright problems such as misappropriation and copying,and the rights and interests of creators are infringed.The existing DIBRoriented 3D video copyright protection schemes are difficult to balance robustness and distinguishability in the face of various attacks.To promote the healthy development of the 3D video market,the research on the copyright protection algorithm of synthetic 3D video based on DIBR is of great significance.Aiming at the problem that the zero-watermarking algorithm is difficult to ensure both robustness and distinguishability under various attacks,especially strong geometric attacks,a zero-watermarking algorithm based on the combination of contrastive learning and Dual-Tree Complex Wavelet Transform is proposed.First,the geometric invariant features of video are mined based on contrastive learning.Then,the second feature is extracted based on the dual-tree complex wavelet transform and singular value decomposition method to resist signal attack and DIBR operation.Finally,the two features are fused based on the attention mechanism,so that the robustness of different targeted features is fully complementary.Experimental results show that the algorithm performs well in the face of strong geometric attacks such as affine transformation,Shearing,flipping,and rotation,and the overall performance is better than existing methods,achieving more comprehensive copyright protection.Aiming at the problem that the zero-watermark scheme has low distinguishability when processing a large amount of video data and the digital watermarking scheme is difficult to resist strong geometric attacks,a digital watermarking and zero-watermarking hybrid algorithm is proposed to eliminate the distinguishability constraint of the zerowatermarking scheme and strengthen the robustness of the digital watermarking scheme to strong geometric attacks.First,the hybrid algorithm inherits the constructed feature of the previous work and stores it in the database and then based on the feature,it further generates a transform coefficient quantization method fused from the owner shared and embedded video,and the digital watermark embedding and extraction method is a Dual-Tree Complex Wavelet Transform.In the process of copyright identification,the feature of the video to be identified is constructed based on the zero-watermark algorithm,and then the original feature is matched in the database.If the matching fails,the noninfringement is identified.If the matching is successful,the matching video is used as the benchmark to calibrate the video to be discriminated based on the Scale-Invariant Feature Transformation method,including rotation and Shearing calibration respectively.Then extract the calibrated and uncalibrated video from the owner sharing,XORed the three extracted results from the owner sharing with the corresponding zero watermark features,fuse the watermark information and complete copyright identification.Experimental results show that compared with the existing schemes,the algorithm significantly improves the overall performance of copyright protection.