Diffusion Weighted Image Reversible Visible Watermarking Algorithm Based on Support Vector Regression

Diffusion weighted imaging is a new technique that was developed based on conventional magnetic resonance sequences. This technique is widely used in clinical medicine. Magnetic resonance imaging is a non-invasive technique that can reflect the proton diffusion motion in human body tissues. Many diffusion weighted images are available on the Internet due to the development of remote diagnosis. However, investigation of academic literature indicates that the copyright information problem for diffusion weighted images has not been studied. To protect the privacy of patients and the copyright information of medical images, we develop a reversible visible watermarking algorithm for medical images via support vector regression. First, a diffusion weighted image is divided into 13 directions, and the b0 direction is selected to embed the watermark. Second, the pixel values of the watermarked image are predicted via support vector regression, and the prediction errors are calculated. Third, a whole location map is generated according to texture classification to increase the watermark robustness. Finally, the relationship between the whole location map and prediction errors is used to process medical images and thus enhance the robustness of the watermarking algorithm. Experimental results show that the proposed algorithm has good robustness. Unknown key information makes the removal of the watermark difficult. Furthermore, the algorithm does not damage medical images and ensures image visibility. These properties can guarantee the integrity of patient information and the authenticity of medical images.