Research On Video Watermarking Methods Based On NMF

As one method for protecting the copyright of video products effectively andpreventing the video products from being tampered with, the video watermark hasattracted more and more attentions. Most of the present video watermark technologiescan resist various and conventional signal processing attacks，however，they are lessrobust to geometric attacks, especially to strong cropping attacks where the embeddedwatermark is cropped enormously. So how to extract and recover the completewatermark in the residual video has always been troubling researchers. At the same time,the experimental research and analysis shows that the existing semi-fragile watermarkhas good fragility but poor robustness, on the basis of deep analysis and research onNMF, this paper propose the following video watermarking methods based on NMF.Firstly, a video watermarking method resistant to strong cropping attacks based onthe non-negative matrix factorization with sparseness constraints on parts of the basismatrix (NMFSCPBM) is proposed. The encrypted watermark is embedded into the bigcoefficients of the basis matrix that the host video is decomposed into by NMFSCPBM.At the same time the watermark embedding strength is adaptively adjusted by the videomotion feature coefficients extracted by the NMFSCPBM method. In watermarkdetection, as long as the residual video contains the numbers of least remainingsub-blocks, the complete basis matrix can be completely recovered throughdecomposition of the non-negative matrix of the least remaining sub-blocks in theresidual video by NMFSCPBM, and then the complete watermark can be extracted. Theexperimental results show that the average intensity resistant to the various regularcropping of this method is up to95.97%and that the average intensity resistant to thevarious irregular cropping of this scheme is up to94.25%. The bit correct rate (BCR)values of the extracted watermark are always100%under all of the above situations. Itis proved that the watermark extraction is not limited by the cropping position and typein this scheme. Compared with other similar methods, the performance of resistingstrong cropping is improved greatly.Secondly, a video watermarking structure based on non-negative matrixfactorization with coefficient matrix normalization (NMFCMN) method is proposed.Copyright protection and content integrity authentication can be achieved at the same time. the host video is decomposed by NMFCMN and the interframe energy differencecoefficient is extracted to construct the zero-watermark. The fragile watermark isformed by embedding the zero-watermark into the least significant bits of thecoefficient matrix. The experimental results show that the zero-watermark is robust notonly to general video processing attacks, such as filtering, adding noise etc, but also tostrong geometric attacks, such as rotation, scaling, translation, cropping, as well ascombined attacks, such as scaling and rotation, rotation and translation, cropping androtation etc, and that copyright protection can be achieved without causing any videodistortion. The embedded fragile watermark is sensitive to tamper with, and can identifythe integrity of the video content accurately. Compared with other similar methods, therobustness and fragility are all improved greatly.