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.