摘要： 在传统的含辅助信息水印检测方案中，通常的做法是从接收到的水印作品中减去原始作品，得到带噪声的水印模型。因此，水印不能完全提取。该文提出了一种基于预测攻击检测水印信号的方案，通过一定的方法预测水印图像所受到的攻击，根据预测结果，对原始图像进行同样的攻击来参与检测。实验表明该检测方案克服了辅助信息的残留噪声，大大提高了水印信息的检出率，改善了整个水印系统的性能。并且，该方案不失一般性，经验证可以应用到其他变换域和各种各样的算法中，有较强的通用性和应用前景。

关键词: 预测攻击, 图像水印, 含辅助信息水印检测

Abstract: In traditional watermark detection scheme, watermark information is extracted by minusing the original media from the received watermarked media. Because of the attack noise, the information can not be completely extracted. This paper proposes a novel detection scheme. It makesa full use of the cover media’s information, determines what kind of attack that the watermarked media has suffered. Then, it attacks the original media using the same kind of attack to higher the relevance between the original media and the watermarked media. In this way, the attack noise can be counteracted when embedded information is extracted with the help of the attacked cover media. Results show that this scheme is especially efficient, and can obviously improve the whole system’s capability. Furthermore, this method can be applied into other domain and various kinds of algorithm, which is promising.

Key words: attack pre-estimation, image watermark, detection with assistant information

中图分类号: 

• TP391