论文部分内容阅读
目的针对当前可逆视频水印隐蔽性和篡改定位能力不足问题,提出一种新颖的基于帧内预测模式的可逆视频水印算法。方法首先,深入分析I帧亮度分量的预测模式对不同类型篡改的敏感性,提取每个帧内编码宏块的预测模式,通过预测模式生成特征码。然后,结合H.264/AVC编码特性和误差补偿算法,在每个亮度4×4残差块中筛选出误差最小系数。最后,运用差值扩展的方法将特征信息作为水印可逆的嵌入到所选系数。结果在含水印视频未受到篡改时,解码端提取水印后可对原始视频进行无损恢复。当视频受到篡改时,算法能精确定位篡改区域并且篡改定位精度达到4×4子块级。由于水印嵌入在误差最小的系数中,能够有效地降低水印嵌入对于视频质量的影响,嵌入水印后图像的PSNR值比现有的基于H.264/AVC可逆水印方案平均提高10%,测试序列的码率增量平均降低了22%左右。结论本文算法较现有算法具有更好的嵌入率、隐蔽性、篡改检测精度,适用于医学、军事、卫星等领域。
Aiming at the problem of the invisible video watermarking and the lack of tampering localization capability, a novel reversible video watermarking algorithm based on intra prediction mode is proposed. Methods Firstly, the sensitivity of the prediction mode of I-frame luminance components to different types of tampering is analyzed in depth. The prediction mode of each intra-frame encoded macroblock is extracted and the prediction mode is used to generate the signature. Then, with the H.264 / AVC coding property and error compensation algorithm, the smallest coefficient of error is screened out for each 4 × 4 luma block. Finally, the method of difference expansion is used to embed the feature information into the selected coefficients as reversible watermark. As a result, when the watermarked video has not been tampered with, the original video can be non-destructively recovered after the decoder extracts the watermark. When the video is tampered with, the algorithm can precisely locate the tampered area and tamper with the positioning accuracy of 4 × 4 sub-block level. As the watermark is embedded in the least error coefficients, the impact of watermark embedding on video quality can be effectively reduced. The PSNR of the watermarked image increases by 10% on average compared with the existing H.264 / AVC based on the reversible watermarking scheme. The average increase in bit rate is about 22% lower. Conclusion The proposed algorithm has better embedding rate, concealment and tampering detection accuracy than the existing algorithms and is suitable for medical, military and satellite fields.