Blind Watermark Extraction Without Reference to the Original Carrier Image

Blind watermark extraction techniques eliminate the need for referencing the original carrier image, thereby broadening their applicability in information hiding and copyright protection. This paper introduces a blind watermarking algorithm based on block-based Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). The method exploits the synchronized variation trends of DCT and DWT coefficients under attacks. By implementing an adaptive embedding strategy that combines relational embedding and quantization embedding, the robustness of blind watermark extraction is substantially improved. Furthermore, dither modulation applied in quantization embedding enables convenient tuning of watermark transparency and robustness. Comprehensive simulation experiments validate the algorithm's exceptional resistance to noise interference, low-pass filtering, JPEG compression, and contrast enhancement attacks, while maintaining high transparency. According to the proposed algorithm, the watermark extraction rate achieves 1.0 when no attack is present, and the Peak Signal-to-Noise Ratio (PSNR) of watermarked images exceeds 45 dB. Even under attacks, the extraction rate remains above 0.9, demonstrating the algorithm's superiority and reliability.