A robust watermarking technique based on image normalization

This paper presents a robust watermarking method against geometric distortions and several other signal processing schemes. In the proposed watermarking method, image normalization is used to get geometric invariant features. The watermark embedding and detection process are carried out in the Discrete Cosine Transform (DCT) domain. A DCT based texture classification is used to classify each image block to embed a stronger as possible watermark, while maintaining watermark imperceptibility. The proposed watermarking scheme is a public watermarking, where the original image is not available for watermark detection. The experimental results show the watermark imperceptibility and robustness against geometric distortion such as rotation, scaling and various other Affine transformation, and common signal processing schemes, such as JPEG compression, Gaussian and impulsive noise contamination.