Graph Fourier Transform Based Image Zero-Watermarking

Abstract

In this paper, we propose an image zero watermarking method using the graph Fourier transform (GFT). We introduce new elements such as sampling suitable for images, convolution of surrounding pixels, and definition of adjacency matrices based on sampling results. To extract features of a host image, a graph signal is produced with an adjacency matrix that depends on the distance from neighboring pixels. Then, the coefficients after the GFT are encoded into binary information using the K-means method. Finally, XOR operation is performed between the obtained features and watermark to achieve zero watermarking. Simulations show that the proposed method works more effectively than a comparative method in resisting a variety of common attacks.