Rate-Distortion Optimized Quantization Index Modulation in Robust Image Watermarking

This paper resents an enhanced scheme utilizing Rate-Distortion Optimized Quantization Index Modulation (RDO-QIM) for image watermarking, emphasizing its ability to balance imperceptibility, robustness and computational efficiency. Unlike conventional QIM-based methods that embed uniformly across all transform coefficients, the proposed approach employs a Lagrangian Rate–Distortion (R–D) criterion to adaptively quantize the coefficients. This dynamically balances embedding distortion and payload rate, achieving superior visual quality without compromising robustness. The algorithm jointly exploits cosine and wavelet coefficients to capture both local and global image features, while a perceptual energy/saliency model directs embedding toward visually tolerant regions. Experimental results on the RSICD dataset and eight standard images confirm significant improvements over a recently published DCT/DWT domains baseline, with PSNR exceeding 50 dB and SSIM above 0.99 on average, alongside a notable reduction in bit error rate under various attacks. Although the method introduces moderate computational overhead (≈25% higher runtime) due to the R-D optimization loop, it maintains full blindness and requires no side information for watermark extraction. These findings demonstrate that RDO-QIM is a highly practical and efficient solution for secure, high-fidelity multimedia watermarking.