We propose a novel scheme for watermarking of digital images based on singular value decomposition (SVD), which makes use of the fact that the SVD subspace preserves significant amount of information of an image, as compared to its singular value matrix, Zhang and Li (2005). The principal components of the watermark are embedded in the original image, leaving the detector with a complimentary set of singular vectors for watermark extraction. The above step invariably ensures that watermark extraction from the embedded watermark image, using a modified matrix, is not possible, thereby removing a major drawback of an earlier proposed algorithm by Liu and Tan (2002).
Download