Fast image watermarking based on signum of cosine matrix

Abstract

In the field of image watermarking, the singular value decomposition has good imperceptibility and robustness, but it has high complexity.  It divides a host image into matrices of U, S, and V. Singular matrix S has been widely used for embedding and extracting watermark, while orthogonal matrices of U and V are used in decomposition and reconstruction. The proposed signum of cosine matrix method is carried out to eliminate the generation of the three matrices at each block and replace it with a signum of cosine matrix. The proposed signum of cosine matrix is performed faster on the decomposition and reconstruction. The image is transformed into a coefficient matrix C using the signum matrix. The C matrix values are closer to the S value of singular value decomposition which can preserve high quality of the watermarked image. The experimental results show that our method is able to produce similar imperceptibility and robustness level of the watermarked image with less computational time.