Lossless medical image watermarking method based on significant difference of cellular automata transform coefficient

Conventional medical image watermarking techniques focus on improving invisibility and robustness of the watermarking mechanism to prevent medical disputes. This paper proposes a medical image watermarking algorithm based on the significant difference of cellular automata transform (CAT) for copyright protection. The medical image is firstly subsampled into four subimages, and two images are randomly chosen to obtain two low-frequency bandwidths using CAT. Coefficients within a low-frequency bandwidth are important information in an image. Hence, the difference between two low-frequency bandwidths is used as an important feature in the medical image. From these important features, watermarks and cover images can be used to generate an ownership share image (OSI) used for verifying the medical image. Besides appearing like cover images, the OSI will also be able to register with a third party. When a suspected medical image requires verification, the important features from the suspected medical image are first extracted. The master share image (MSI) can be generated from the important features from the suspected medical image. Lastly, the OSI and MSI can be combined to extract the watermark to verify the suspected medical image. The advantage of our proposed method is that the medical image does not require alteration to protect the copyright of the medical image. This means that while the image is protected, medical disputes will be unlikely and the appearance of the registered OSI will carry significant data to make management more convenient. Lastly, the proposed method has the features of having better security, invisibility, and robustness. Moreover, experimental results have demonstrated that our method results in good performance.