The anti-geometric attack digital watermarking algorithm based on image normalization of local information

Objective: On the issues of the insufficient embedding capacity of most existing anti-geometric attack watermark algorithms, and the disability to shear attacks, an algorithm using lo-cal information which based on image normalization and Gaussian-Hermite invariant moments is proposed. Method: In this algorithm, a local area selection strategy based on integrity authentication is presented to achieve exact area matching before and after the attacks; then, on the basis of the good geometric stability of Gaussian-Hermite invariant moments, characteristic matrix is structured, through which the watermark embedding and extraction is realized. In order to expand embedding capacity, an improved local normalization strategy is put forward on the combination of local area selection and normalization. Result: Experiments show that our algorithm has good robustness against rotating, translation, scaling and can resist shear attacks up to a certain degree.