Enhancing image watermarking with blockchain: deep residual network and Addax Mayfly Optimization for optimal region selection

Secure medical image watermarking is essential for protecting image integrity, ownership, and authenticity during healthcare data exchange. However, existing watermarking approaches often face limitations in selecting suitable embedding regions and maintaining watermark robustness against image-processing attacks. These limitations highlight the need for enhanced watermarking techniques that improve region selection, image quality, and data security. To address these challenges, this research proposes an Addax Mayfly Optimization-based Deep Residual Network (AMFO_DRN) for intelligent region-of-interest (ROI) selection. The framework employs Quaternion Curvelet Transform (QCT) for multi-frequency feature extraction, while a Deep Residual Network (DRN) identifies suitable embedding regions. Addax Mayfly Optimization optimizes DRN parameters to improve ROI selection. Blockchain and InterPlanetary File System (IPFS) provide immutable ownership records, decentralized storage, and integrity verification. Experiments using the TB Chest X-ray dataset achieve SSIM, NCC, and PSNR values of 0.877, 0.908, and 35.877 dB, respectively.