A novel color image watermarking scheme based on two-stage partitioning and 3D-QIM

Abstract With the rapid advancement of 5G networks, the pace of content creation in the digital media industry continues to accelerate, leading to a growing demand for copyright protection. However, the conventional one-dimensional quantization index modulation (1D-QIM) watermarking scheme suffers from a limited payload and relatively weak resistance to typical signal processing operations. To address these issues, this paper proposes a blind color-image watermarking algorithm based on a two-stage partitioning strategy in the Discrete Cosine Transform (DCT) domain combined with three-dimensional quantization index modulation (3D-QIM). The algorithm employs a secondary partitioning of pixel blocks, followed by direct computation of the DC coefficients of the resulting sub-blocks in the spatial domain. It then utilizes 3D-QIM to establish an efficient and robust watermarking mechanism. Furthermore, the Dreaming Optimization Algorithm (DOA) is introduced to optimize experimental parameters, ensuring an optimal balance in performance. Experimental results demonstrate that the proposed algorithm exhibits outstanding invisibility and robustness, performs reliably under various attacks, and meets real-time application and security requirements. Specifically, it achieves PSNR values all exceeding 40 dB, an average Structural Similarity Index Measure greater than 0.96, an average normalized correlation (NC) value above 0.95, and a total embedding and extraction time of just 0.2369 seconds.