Robust zero-watermarking algorithm via multi-scale feature analysis for medical images

Abstract

With the rapid growth of information technology, the development and implementation of copyright protection for medical images has become crucial. In this paper, we develop a distinguishable zero-watermarking algorithm via multi-scale feature analysis for medical images. We first detect the global features of the image with speeded-up robust features (SURF) and select the feature regions from the image through texture analysis. Then, we adopt local binary pattern (LBP) to detect the local texture features of these feature areas, and perform singular value decomposition (SVD) to extract the scale features and the detail features; these features are fused to form the feature matrix, and the average hash (aHash) algorithm is applied to the feature matrix to generate the binary feature map. Finally, we perform exclusive-or (XOR) operation between the feature images and the watermark image to generate zero-watermarks, which will be stored in the copyright protection center for further copyright authentication. Experimental results show that the average NC value of the proposed algorithm reaches 0.99 under most attacks, and the average BER of similar image extraction watermark keep below 0.27, which outperforms the current state-of-the-art (SOTA) watermarking algorithms.