DFT-Based Reversible Watermarking Method for Image Ownership Protection

Abstract

The Performance of Discrete Fourier Transform (DFT)-based watermarking methods has been carefully examined in the literature. Although the watermark in most of the literature was embedded in the DFT magnitudes using bit plane embedding, it was recently embedded in the Direct Current (DC) coefficient in the spatial domain. However, data loss due to rounding and replacement operations are still evident. Therefore, the objective of the method proposed here was to combine previous literature designs to implement a reversible DFT-based watermarking method for image ownership protection using bit plane embedding in the DC coefficient. The watermark was embedded in a middle bit plane of the DC coefficient for each DFT-transformed image block. In order to ensure reversibility and improve the security level, a combination of double Feynman and XOR gates was used to shuffle the watermark bits. The results revealed that the 8th (PSNR/SSIM = 32dB/0.8826), the 9th (PSNR/SSIM = 38dB/0.0.9587), and the 10th (PSNR/SSIM = 44dB/0.9917) bit planes for block sizes of 4×4, 8×8, and 16×16, respectively, were the best bit planes showing good imperceptibility and resistance to compression, filtering, and noise attacks. In conclusion, embedding the DC coefficients rather than all the magnitudes has influentially increased the watermarking robustness. In contrast, embedding the DC coefficients in the frequency domain rather than the spatial domain reduced the image’s structural contents distortion. Furthermore, the proposed method for grayscale images is effective in applications where reversibility is desired. However, further studies to find colored images’ reversible methods are recommended.