Study Of Video Watermarking Resisiting To Special And Temporal Desynchronization Attacks

Digital watermarking is some labels or signatures which are embeddedinto digital multimedia, then carries out the function such as informationhiding, storing, copyright safeguarding by tracking the use of the digitalmultimedia by detecting the watermarks. At present, watermark technologybecomes a kind of effective means to multimedia copyright authenticationand its integrity safeguarding.Anti-attack ability is one of the key restricted factors for applicationsof digital video watermarks. In the practical environment, the geometric andtemporal distortions are commonly used to invalidate the videowatermarking intentionally or unintentionally. As a countermeasure, a novelvideo watermarking that can resist the spatio-temporal desynchronizationssimultaneously is presented in this paper. The resistance of the novel videowatermarking to the temporal distortion is also proved in the experiments.First of all, according to the principle of the algorithm alreadyproposed, this paper presents a novel video watermarking that can resist thegeometric attacks. In the proposal, we split the video test sequence intoframes, and then divide each frame into several 8*8 size blocks, do thediscrete cosine transformation (DCT) on these blocks, after taking out eachDCT transformation, we keep the direct current coefficient(DC coefficient).According to the experiment research, we found that, though there is noregulation in the variety of the single DC coefficient in the transformdomain, the energy of the DC coefficient will transfer to the coefficients?ˉthat around during the geometric distortions such as rotation, scaling etc.The feature of the transform domain that will not be altered easily with thegeometric distortions is investigated, and is defined as follows:The experimental results show that the curve change little during therotation transformation, the scaling transformation just change the curve?ˉsdisplacement, and all of this will not influence the result of the relatedde-mean watermark detection. Then take advantage of the feature of thetransform domain which will not be altered easily by the geometricdistortions, we proposed a novel video watermarking that can resist thegeometric attacks. The watermark information is embedded via altering thisfeature along the temporal axis with the help of the 127-bit sequence codesof spread-spectrum. The particular watermark embedding and detectionprocedures are as follows: According to the length (N) of thespread-spectrum sequence codes, the embedder will select a N-framesub-sequence of the testing video, this N-frame sub-sequence will be anembedding period. In every period, do the following operations:For each 8*8 block, do the discrete cosine transform, then take outdirect current coefficient(DC coefficient) in each block, using thetransformation above for each DC coefficient, the gold(i, k) stands for thecode of the spread-spectrum we use when embedding the differentwatermark bit in different frames, after all of the watermark informationwhich we want to embed is embedded, then do the inverse discrete cosinetransformation (IDCT), and store the current frame in a new file. Thecorresponding watermark detection process is as follows, according to thelength of the sequence codes of spread-spectrum, the related detector needsto operate on the corresponding N-frame of the video testing sequence. Weuse the formula below:If the |?d we calculated is bigger than the threshold value we defined inadvance, it shows that the watermark information can be detectedsuccessfully, the value is determined by the sign of the |?d.Although this algorithm can resist the geometric attacks effectively, itcan't be resilient against temporal attack. The reason is that after insertion,dropping attack, the frame ordinal number change, the detector andembedder use the different sub-sequence. Then we introduce another videowatermarking proposal, in this algorithm, it describes a method for efficienttemporal synchronization and resynchronization for video watermarking.Taking advantage of the temporal redundancy, the embedded watermark canbe resilient against temporal attack such as frame dropping, insertion,transposition, and averaging, but to the geometric distortions, thesynchronous information imbedded before will lose completely, then wecombine the two method mentioned above, proposed a novel videowatermarking that can resist both geometric and temporal distortions, thenew watermark embedding and detection protocols are shown below: Westart the new algorithm on the basis of anti-geometric attack protocol, usethe first frame in every period as the synchronous frame, only in this kind offrames, we do the synchronous embedding work. In this way, although thesynchronous information in the synchronous frame missing, the anti-geometric attack algorithm can still detect the watermark informationsuccessfully without any influence after the geometric distortion;When itcomes to the temporal distortion such as frame dropping, insertion etc,though some of the ordinal frame numbers change, we still can use thesynchronous information in the synchronous frame to achieve theresynchronization, and detect the watermark information successfully.Simulations are performed to evaluate the performance of the proposedalgorithm. We selected test results from several representative video testingsequence, the experimental results show that the proposed method is robustagainst geometrical and temporal distortions.