Image-adaptive watermarking using maximum likelihood decoder for medical images

In this paper, a new medical image-adaptive watermarking technique is proposed in which embedding of the watermark is done in low frequency coefficients for achieving high robustness using an adjustable dynamic strength factor. The low frequency coefficients are modeled using Gaussian distribution to design a Maximum Likelihood (ML) decoder. The decoder recovers the watermark with the help of side information containing the adjustable dynamic strength factor, position of blocks (used for embedding), mean and variances of low frequency coefficients. The method contributes towards a highly flexible and easily adjustable dynamic strength factor for achieving the best imperceptibility with the highest robustness. The validity of the new technique is verified against various attacks and the results are compared with other watermarking schemes. The proposed technique is found to generate better results.