Study On H.264/AVC Video Authentication Technology Based On Fragile Watermarking

With the development of Internet and computer technology as well as the wide application of the video encoded by H.264/AVC, the people need to do the H.264/AVC video content integrity authentication. Fragile watermarking technology does not allow any tampering to the certification object, so it can be used for content integrity authentication. Therefore, the study of H.264/AVC fragile watermarking technology is very meaningful.This paper presents and improves two algorithm of H.264/AVC fragile watermarking used for content authentication; the main work is as follows:As the first algorithm, a synchronization watermark is embedded into the current GOP group among the P and B frames, which is generated by a stream cipher encryption device whose inputs are a key and the Hash value of the reconstructed I-frame. The basic embedding idea is to modify the number of trailing coefficients of a16×16macroblock according to the embedded information, combining Tian’s and Zhao’s algorithm can implement that each5-bit watermark information is embedded by modifying a4×4sub-macroblock tailing coefficients at most. Experimental results show that the distortion drift can be reduced by embedding the watermark information to the P and B frames; this algorithm can’t only protect the full content of the I-frames, but also can have the full capabilities of identifying erasure frames, increasing frames, and can have a little of positioning capability to the P and B frames’contents tampering; Embedding watermark will not bring greater visual distortion and bit-rate increases to the video.In order to enhance the authentication capability to the P frames and B frames as well as avoid distortion drift. This paper has proposed a reversible fragile watermarking algorithm for H.264/AVC video. Firstly, the features of one group of reconstructed video frames before filtering are determined, according to the center of mass of the histogram characteristic function (HCF). Secondly, the unique watermark for authentication, which is derived from features, an equipment number and a timestamp, is embedded into the quantized Discrete Cosine Transform (DCT) coefficients of residual luminances of the next1-frame by a reversible algorithm. The embedding basic idea is that we modify the number of the4×4sub-macroblocks’coefficients to the maximum in the macroblock, so that the factors of the4×4watermarked sub-macroblock have two characteristics:First, the number of non-zero coefficients is the most and the last coefficient is-1. The second is that the last factor is largest in the absolute last factor of all macro blocks. We can easily implement the video authentication by comparing the restructured authentication information with the extracted one. Experimental results show that our watermarking can be successfully embedded and correctly extracted without introducing visual artifacts. Moreover, the original video content can be completely restored after extracting the watermarking from the watermarked-video. Our algorithm can resist on the intra frame content tampering, frame deletion, frame insertion, frame exchange, recompression, and conventional signal processing, can maintain a good trade-off between video visual quality and embedding capacity. In addition, our algorithm can keep low bit-rate-difference, and can authenticate the integrity of all color components of video frames.