Abstract: A novel digital watermarking algorithm based on singular value decomposition(SVD) was proposed. The data was collected in amplitude of image frequency domain through setting the start parameter and termination parameter of sampling. Data matrix for embedding watermarking was constructed by using these collection data. The algorithm made SVD on blocks of data matrix, and got a sub-maximum singular value sequence. Watermark was embedded via modifying the sub-maximum singular value. The modification method is a additive means,and intensity can be adjusted according to the image characteristics. The watermarking algorithm can resist geometric distortion depending on data matrix construction method and geometric attacks invariance of sub-maximum. I}he experimental results show that the method is more flexibility and robustness than traditional methods.

Key words: SVD, Digital watermarking, Gcometric distortion