提出了一种基于图像内容的二次水印的理论模型。通过提取能够唯一标识原始图像的信息摘要MD5，将其和图像水印依次嵌入到原始图像中，使得嵌入的水印具有与原图相关和不相关的双重意义；通过有实际内容的图像水印来实现水印系统的鲁棒性；通过MD5与原始图像的关联性来抵抗解释攻击。实验结果表明，提出的方法简单可行，内嵌数据量小，与图像文件关联性强，具有较强的鲁棒性。

A quadratic watermark theoretic model based on image content was proposed. The model extracted MD5 value firstly, which is the unique identification of information abstract of the origin image. Then the MD5 value was embedded into the origin image together with image watermark, which makes the embedded watermark include both contents relevant and irrelevant to the content of the origin image. The robustness of the watermark system was realized by using image watermark of practical content; the explanation attack was resisted with the association of MD5 value and the origin image. Experiments results showed that the method proposed is simple, feasible, low embedded data flow, high relevancy to the image file, and high robustness.