This paper proposes a method for experimentally estimating the distribution of correlation in correlation-based digital image watermarking. The proposed method evaluates how distortions affect the correlations between image and embedded and non-embedded watermarks. The method involves a computational experiment using an image dataset, where correlations are measured before and after filtering for embedded and non-embedded watermarks. Applying the proposed method to a DCT based watermarking algorithm demonstrates that filtering significantly reduces the correlation with the embedded watermark, while having minimal changes for non-embedded.
Информационные технологии и математическое моделирование (ИТММ-2025) : материалы XXIV Международной конференции имени А. Ф. Терпугова, 20-25 октября 2025 г.. Томск, 2025. С. 78-83