Digital image watermarking in sparse domain

A watermarking method based on a robust sparse domain is proposed in this paper, which integrates the secret information into the significant sparse elements of the original image. Our algorithm protects the original data by a two-way security process to embed confidential information. First of all, converting the watermark logo into a discrete transform coefficient (DCT) is the protection process. Then, using the dictionary learning method, the transformed coefficient is embedded in the selected effective sparse coefficient in the original image. The embedded logo is extracted from the selected effective sparse coefficient using the sparse orthogonal matching tracking algorithm (OMP) domain. Then, the discrete inverse transformation is performed. To check the algorithm's efficiency, numerous specific attacks are checked. The experimental results show that the algorithm can recover the embedded watermark with precision without losing any information.