为了提高水印算法的鲁棒性，安全性，提出了一种基于图象内容的水印模型（CBWM：Content-Based Watermarking Model)，其基本原理是水印向量被嵌入到图象的特征向量中，且两者相互正交，在此模型中，静止图象的水印检测被归结于Neymann-Pearson准则下的统计模型，将该模型应用于通用的DCT水印算法，从实验结果看，算法性能有很大地提高。

In this paper, we proposed a new general additive watermarking model based on the content of digital images, called as CBWM (Content Based Watermarking Model). It provides a common basis to study many existing watermarking algorithms that are based on unitary transforms and to evaluate their performance and characteristics. CBWM is designed to address two important issues of digital watermarking. One is the requirement of robustness. In order to improve robustness and security, the embedded watermark is designed to be orthogonal to the feature vector of the original image, which means that watermarking casting is image content dependent. The second issue is watermark detection. CBWM presents a statistical approach to watermark detection based on the Neymann Pearson criterion and describes a method for computing the probability of false alarm and missing rejection during watermark detection. In experimental tests, CBWM is applied to the popular discrete cosine transform and promising results are obtained.