An Image Watermarking Algorithm in DCT Domain Based on Optimal Patterns

P. 1–5.

Anna Melman, Oleg Evsutin, Danil Smirnov

Sharing images via social media and specialized sites creates a copyright issue. Image watermarking methods provide copyright protection for authors and owners of digital content. The security level of a watermarking algorithm depends on a watermark’s resistance to various distorting effects, such as brightness changing, contrast changing, applying a Gaussian filter, and others. At the same time, a watermark needs to be invisible in most cases, so as not to disturb a perception of an image by a viewer. In this paper, we propose a new image watermarking algorithm in a domain of discrete cosine transform, which provides a balance between embedding imperceptibility and robustness. This algorithm uses redundant embedding based on patterns and synchronization blocks to provide embedding robustness to image processing operations. A high level of imperceptibility is provided by optimizing patterns using a genetic algorithm. Experimental results demonstrate high values of key quality metrics such as PSNR, SSIM, and BER.

Keywords: genetic algorithm, image processing, watermarking, discrete cosine transform

Publication based on the results of: Robust digital data authentication methods for information systems (2023)

In book

2023 XVIII International Symposium Problems of Redundancy in Information and Control Systems (REDUNDANCY)

IEEE, 2023.