Blind Watermarking for Hiding Color Images in Color Images with Super-Resolution Enhancement

Current watermarking techniques often struggle to maintain both high robustness and perceptual quality when embedding and extracting color images. Prior research has shown that watermarking in the discrete cosine transform domain can offer some advantages, but limitations persist in resisting common attacks. Additionally, the visual quality of extracted watermarks remains a challenge. This gap motivated the development of a new method that integrates super-resolution enhancement. No prior work had resolved the issue of combining robust watermarking with perceptual improvements. Existing approaches tend to sacrifice either robustness or visual fidelity. The need for a dual-purpose solution became evident. This paper introduces a novel approach that addresses both concerns. The study builds on established watermarking principles but introduces a unique modulation method. The integration of generative adversarial networks marks a departure from traditional post-processing techniques.

The aim of this study is to develop a robust and visually effective method for embedding and extracting color watermarks in color images. The specific problem addressed is the trade-off between watermark robustness and perceptual quality. Traditional watermarking methods often fail to maintain high fidelity after attacks. The motivation stems from the need for secure and visually acceptable watermarking in digital media. The proposed method seeks to overcome these limitations by modulating discrete cosine transform coefficients. Additionally, the study introduces a generative adversarial network to enhance the extracted watermark. The goal is to achieve high zero-normalized cross-correlation scores while preserving visual quality. The approach is designed to outperform existing watermarking schemes in real-world scenarios. The integration of super-resolution reconstruction is intended to improve perceptual quality further.

The proposed method embeds a color watermark by modulating the gap between paired coefficient magnitudes in the discrete cosine transform domain. This modulation aligns the gap with the intended pixel value of the watermark. Watermark extraction involves regaining and regulating the gap distance to recover the original intensity values. To assess robustness, the method was compared against seven existing watermarking schemes. The evaluation used zero-normalized cross-correlation as the primary metric. A generative adversarial network was introduced to enhance the visual quality of the extracted watermark. The GAN performed image denoising and super-resolution reconstruction. The study tested the method against common attacks such as cropping, filtering, and compression. The results were analyzed to determine the effectiveness of the proposed approach.

The proposed watermarking scheme outperformed seven existing methods in terms of zero-normalized cross-correlation. The method achieved ZNCC scores higher than 0.65 in most attack scenarios. The only exception was JPEG compression, where performance slightly declined. The generative adversarial network significantly improved perceptual quality. The GAN-enhanced watermarks had ZNCC scores of no less than 0.78. These results indicate a strong correlation between the extracted and original watermarks. The method demonstrated robustness against cropping and filtering attacks. The integration of the GAN provided a noticeable enhancement in visual fidelity. The study confirmed that the proposed approach maintains high robustness and quality. The results suggest that the method is suitable for real-world applications.

The authors propose that the new watermarking method effectively balances robustness and perceptual quality. The modulation of discrete cosine transform coefficients allows for precise watermark embedding. The use of a generative adversarial network enhances the visual quality of the extracted watermark. The method outperformed seven existing schemes in most attack scenarios. The results suggest that the approach is suitable for practical applications. The study highlights the importance of integrating super-resolution techniques. The authors suggest that the method's performance is particularly strong against non-compression attacks. The findings support the use of the proposed scheme in digital media security.