Hierarchical multilevel framework using RDWT-QR optimized watermarking in telemedicine

摘要

A hierarchical multilevel medical image watermarking framework is introduced in this paper. First, robust watermarking is performed to verify the copyright of the ownership. Later, an improved fragile watermarking is implemented on the robust watermarked image to achieve tamper localization along with self-recovery capability. The robust insertion is also optimized with the Particle Swarm Bacterial Foraging Optimization algorithm to ensure improved robustness and imperceptibility metrics of watermarking. During robust watermarking, multiple watermarks are concealed in the first column of upper triangular matrix 'R,' of the host image, transformed by a hybrid combination of Redundant Discrete Wavelet Transform-QR. This hybrid combination exploits the redundancy in Redundant Discrete Wavelet Transform to improve the payload capacity, imperceptibility and the low computational QR transform is used to make watermarking scheme free from the false positive problem. First, the cover image is fragmented into blocks of 2 x 2 dimension, and then, on each block, QR decomposition is carried out. The upper triangular matrix pattern is utilized to figure out the tamper localization bits for each block. Fragile watermarking is implemented on the robust watermarked image, in which the watermark is formed by encoding all 2 x 2 blocks with QR coefficients which are hidden into blocks reserved for self recovery. Authentication of watermark bits, which is different for each watermarked image, helps in resisting the block-based attacks such as Vector Quantization, Collage and Constant Average Attacks. In comparison with the existing works in literature, the proposed framework offers high average Peak Signal-toNoise Ratio and Normalized Cross Correlation in the robust watermarking. In addition to this, excellent results is achieved in the fragile watermarking with an overall tamper detection rate of 99.7% and an overall recovery of 57 dB Peak Signal-to-Noise Ratio for a 50% tampering.