Cauchy Distribution Based Data Hiding And Its Detection For H.264/AVC Video Stream

With the rapid development of digital television, video surveillance systems and video streaming, the protection of copyright and privacy information has gained more and more attention. Information hiding is an important branch in the field of information security, which provides a new solution for the copyright protection of digital products.As a new video coding standard,H.264/AVC can achieve higher compression ratio and better network friendly. Therefore,the research on information detection and its detection for H.264/AVC video is of significant importance.After a brief introduction to the concepts,characteristics and classifications of information hiding techniques,the thesis makes use of the characteristics of H.264/AVC video coding standard, and puts emphasis on the stastically distribution of the transform coefficients,and DCT coefficients based information hiding and detection techniques.The main work is summarized as following:Firstly, Four statistical models,which include Laplace model,Generalized Gaussian model (GGD),a model and Cauchy model,are investigated.Two models, i.e.,GGD and Cauchy model are compared using chi-square goodness of fit test and the K-S test. By experimental comparison, it is verified that Cauchy distribution model is optimal to describe the distribution of the DCT coefficients for H.264/AVC.Secondly, a Cauchy distribution based watermark detection method is proposed for H.264/AVC video stream.It introduces the concept of ternary hypothesis testing, and the watermark detection is turned into a problem of hypothesis testing.A Watson visual model based video watermarking algorithm,which embeds bipolar additive watermark into the quantized residue of DCT coefficients,is used as example for experiment and performance analysis.Experimental results demonstrate that the proposed approach can achiever detection ratio of 80 percent.For other watermarking algorithms with high energy, it still can obtain satisfactory performance.Thirdly, a novel privacy information hiding algorithm is proposed for H.264/AVC,so as to protect the privacy information in video surveillance system. Privacy information is firstly extracted and binarized as secret data to be embedded into video.Embedding is performed by finding the edge and texture regions for the DCT coefficient residues and intra-coded Y, Cb,Cr components.Experimental results show that the proposed approach is of large embedding capacity, low noise and no requirements for watermark energy. For the detection of privacy information,binary hypothesis testing based on Cauchy model is used.Experimental results demonstrate that the false detection ratio of Cauchy detector is lower than that of GGD detector and a detector.