Adaptive Watermarking Scheme for Enhanced Image Authentication and Restoration

Traditional fragile watermarking schemes often struggle to balance tamper detection accuracy, image quality, and processing speed. This paper presents a novel adaptive fragile watermarking scheme that leverages Canny-edge detection to dynamically adjust watermark embedding based on image features. The proposed approach classifies image regions into smooth and dense areas. Larger 4×4 blocks are employed in smooth regions, prioritizing visual quality and processing speed, while smaller 2×2 blocks are used in edge and dense regions to enhance tamper detection accuracy. This adaptive strategy effectively balances competing factors, resulting in superior recovered image quality compared to fixed-block schemes. Experimental results demonstrate the effectiveness of the proposed scheme. Watermarked images exhibit an average PSNR of 46.5 dB, maintaining high visual quality. The scheme achieves over 97% tamper detection accuracy even with significant tampering exceeding 90%. Furthermore, the proposed method demonstrates robustness against collage attacks and outperforms existing techniques in both tamper detection and visual quality.