一种抗重压缩的H.264/AVC视频水印算法*

针对当前视频水印抗重压缩能力不足,提出一种基于I帧色度量化残差系数的视频水印算法。通过对H.264/AVC帧内预测模式的分析,得出重压缩后预测模式的转变是造成水印提取失败的主要因素,因此,算法选择在更加稳定的色度分量中嵌入水印。首先,利用4?4子块中非零系数个数选择合适的色度子块,然后,运用差分能量的思想,结合Arnold置乱后的水印信息,通过调制其中一个非零量化DCT残差系数改变两个特定位置系数的关系嵌入水印。算法在帧内预测模式色度分量中嵌入水印,有效的解决了重压缩后水印提取不同步问题,提高了水印的抗重压缩性能。实验结果显示,提取的水印图像位错误率(BER)在20%以下,重构图像的SSIM比现有方案提高了1.2%-4.4%,另外,测试序列的比特增加率(BIR)平均降低了70%左右。算法适用于广播监控等实时视频应用。

A H.264/AVC video watermarking algorithm against recompression

In order to solve the shortage of video watermarking ability to resist recompression attacks, a video watermarking algorithm is proposed based on chrominance quantized residual coefficients of I frame. Through the analysis of H.264/AVC intra prediction mode, it is concluded that the main factor which results in failure of the watermark extraction is prediction mode transformation generated by recompression. Therefore, the more stable chrominance component is chosen to embedding the watermark in proposed algorithm. Firstly, the proposed algorithm utilizes the number of non-zero coefficients of 4?4 blocks to select candidate blocks. Then, According to thought of differential energy, the watermarking information scrambled by Arnold transformation is embedded by slightly modulating one DCT residual coefficient of the two specific non-zero quantized DCT coefficients. Embedding the watermark in chrominance component has effectively solved the asynchronous problem during watermark extraction owing to recompression, as well as improving the watermarking resisting capacity against recompression. The experimental results show that bit error rate (BER) of the extracted watermark image is below 20% and SSIM of reconstruction image is improved by 1.2%-4.4% compared with previous scheme. Furthermore, the bit increase rate (BIR) of test sequence decreased by an average of 70%. The proposed algorithm is suitable for real-time video applications such as broadcast monitoring.