基于块差值的半脆弱可逆视频认证方法

Video Content Authentication Based on Invertible Semi-fragile Watermarking

针对半脆弱可逆视频认证问题,提出一种基于块差值直方图调整的半脆弱可逆视频认证方法。在H.264帧内预测后重构的I帧Y分量中,通过预测残差对图像块分类。根据人眼视觉模型,在纹理复杂的区域嵌入比纹理平滑区域更多的水印。在接收端使用K-means对各块的差值进行聚类,自适应判断块差值所属区间,实现高准确率的水印提取。通过空间和时间认证之后,通过嵌入算法的逆操作实现无损视频还原。实验结果表明:提出算法具有良好的视觉效果,能够抵抗常规图像处理和噪声攻击,并实现对恶意篡改区域定位功能。

An invertible semi-fragile watermarking method based on the difference of image blocks is proposed for lossless video content authentication.After I frame H.264 reconstruction,H.264 prediction residuals will help classify image blocks.According to human vision system,more watermarks will be embedded into texture areas than smooth ones.At the receiver side,K-means is used to cluster the difference to achieve high-accuracy watermark extraction.After spatiotemporal authentication,the original frame will be recovered.Experimental results show that the proposed method has good invisibility.And it is sensitive to intentional malicious attacks while it is robust to legitimate manipulations,such as certain level of lossy compression,channel noise and Gaussian filtering.