부산대학교 도서관

Authentication and confidentiality of a patient's data records is an extremely important issue in digital store or telehealth services. Authentication solutions based on the watermarking method have been widely used in image processing fields due to its simplicity. However, the watermarking process introduces noise into the covered image that can affect the medical image quality. To overcome that, this paper proposes a curvelet-based authentication method to keep the medical image quality in certain watermark information. Moreover, the proposed scheme uses a machine learning method to recover the patient's record information in extracted watermark data. Comparative numerical results show a certain effect on image quality after embedding and the ability to recover embedded information when compared with the discrete wavelet transformation method.