A cross-embedding based medical image tamper detection and self-recovery watermarking scheme

With the rapid growth in communication and computing technologies, the transmission of medical images over the Internet is on the rise. In such a scenario, there is a special need to meet the security and privacy issues and challenges of individual and Intellectual Property (IP) owners. It is highly important for an individual to keep his/her personal medical images against invalid manipulation by impostors. Hence developments of authentication and tamper detection techniques are the need of the hour. For this, a tamper detection and self-recovery watermarking scheme for medical images based on texture degree and cross-embedding is proposed in this paper. Firstly, divide medical images into ROI (Region of Interest) and RONI (Region of Non-Interest); generate a double authentication watermark in ROI to improve the accuracy of tamper detection and reduce the probability of false alarm; calculate texture complexity based on 4-dimensional features in ROI, and divide ROI into texture blocks and smooth blocks; generate different recovery watermarks according to the characteristics of different blocks using compression-aware technology. Then, hide the recovery watermark in RONI based on the reference matrix and cross-embedding technology. Finally, locate the tampered blocks in the ROI based on three level tamper detection strategy including pixel-level, block-level, and multi-direction subband-level; restore the tampered region by the extracted recovery watermark. The experimental results indicate that the tamper detection accuracy of the ROI region is close to 100%. Additionally, at an embedding rate of 1.4074bpp, the PSNR reaches 45.0217 dB and the NC is 0.99. In addition, the scheme provides promising results against copy-paste attacks, collage attacks and steganalysis. Also, the scheme achieves privacy protection. This clearly demonstrates that the proposed scheme has several advantages, including strong tamper detection capability, effective self-recovery, high security, excellent concealment, and robustness.