A robust zero-watermarking algorithm for color image based on tensor mode expansion

Digital zero-watermarking technology is an effective measure to protect image copyright, and many algorithms have been proposed based on zero-watermark. However, most of the existing zero-watermarking algorithms are designed for grayscale images. In this paper, a new zero-watermarking algorithm for color image based on tensor mode expansion is proposed. In the proposed scheme, four images of R, G, B, and gray are firstly generated from the original image. Then, the four images are appropriately combined to construct two three-dimensional tensors, and tensor expansion is performed on the obtained tensors. By performing singular value decomposition (SVD) and discrete cosine transform (DCT) on the expanded data, a feature image containing the main information of the host image can be generated. Finally, the feature image is fused with the specified logo image to obtain a zero-watermark image. Experimental results show that the proposed algorithm is robust to noise attacks and common image processing attacks, and better performances compared with some existing algorithms are also achieved. In addition, since the proposed algorithm is based on color images, it can make full use of all aspects of color images.