Robust zero-watermarking method for medical images based on FFST and Daisy descriptor

With the continuous development of digital medical imaging technologies, ensuring the security of the medical images has become critically important. In this study,the Daisy descriptors’ stability against attacks was first experimented with, and the findings show that it provides superior robustness. With this, a robust zero-watermarking method is designed to maintain medical image integrity and enable copyright protection by combining the fast finite Shearlet transform (FFST), Daisy descriptor, and Hessenberg decomposition. First, FFST was performed on the medical image to extract the low-frequency component and divide it into blocks of equal size. Second, each block’s Daisy descriptor matrix is calculated and its 8$\times$ 8 block is selected, after which the Hessenberg decomposition is performed for each block, and a feature image is derived from the magnitude comparison between the maximum value of each block and the global mean. Additionally, the copyrighted image is first encrypted by using a 2D Logistic-Sine coupling mapping, and then combined with the feature image through an exclusive OR operation to produce an unrecognizable binary image. The experimental results on ten medical images and three benchmark image databases (COVID-19, OASIS-1, and SIPI) show that the proposed method is highly resistant to most attacks, and the normalized correlation coefficient is always maintained higher than 0.95. Compared to typical methods, our method achieves superior robustness and improves the average performance by approximately 3.2%.