Blind video watermarking scheme for medical video authentication.

Medical video watermarking is one of the beneficial and efficient tools to prohibit important patients' data from illicit enrollment and redistribution. In this paper, a new blind watermarking scheme has been proposed to improve the confidentiality, integrity, authenticity, and perceptual quality of a medical video with minimum distortion. The proposed scheme is based on 2D-DWT and dual Hessenberg-QR decomposition, where the input medical video is initially processed into frames. Then, the processed frames are transformed into sub-bands using 2D-DWT, followed by applying Hessenberg-QR decomposition on the selected wavelet HL2 sub-band. The watermark is scrambled via Arnold cat map to raise confidentiality and then concealed in the modified selected features. The watermark is extracted in a fully blind mode without referencing the original video, which reduces the extraction time. The proposed scheme maintained a fundamental tradeoff between robustness and visual imperceptibility compared to existing methods against many commonly encountered attacks. The visual imperceptibility has been evaluated using well-known metrics PSNR, SSIM, Q-index, and histogram analysis. The proposed scheme achieves a high PSNR value of (70.6899 dB) with minimal distortion and a high robustness level with an average NC value of (0.9998) and BER value of (0.0023) while conserving a large payload capacity. The obtained results show superior performance over similar video watermarking methods. The limitation of this scheme is the elapsed time during the embedding process since we utilized dual Hessenberg-QR decomposition. One possible solution to reduce time consumption is simple decompositions like bound-constrained SVM or similar decompositions.