Medical image watermarking encryption algorithm based on spiral transform and shift registers

The popularity of telemedicine has opened up the possibility of realizing scenarios such as multi-party online consultations and remote surgical protocols for treating patients. In the process of diagnosing and treating patients, medical image movies containing the patients' physical conditions are transmitted across regions. Given that the security of medical image transmission is directly related to patient safety, this paper aims to address encryption issues in this field. A novel watermark encryption scheme is proposed, which adopts a hybrid architecture of dynamic spiral operations and shift registers. In obtaining the embedded watermark, focusing on the non-visualization and time cost, the edge features of medical images are first extracted by the Canny operator and are used as the watermark information. In the specific embedding implementation, NSCT, DCT, and SVD techniques are combined to realize a good embedding of the watermark. After obtaining the medical watermark image, the multi-size watermark image is merged and converted into a cube of plaintext information. Chaotic sequences generated by coupling discrete memristors with Rulkov neurons are used to define the starting point and direction of spiral operations, thereby disrupting the pixel positions within the cube. The diffusion part utilizes the shift register principle to convert the pixel values into 8-bit strings. This is followed by a dissimilarity operation with the chaotic sequence, thereby altering the pixel values to produce the final ciphertext cube. The scheme successfully passed a series of tests, including key space and key sensitivity, histogram, pixel correlation, noise attack, etc. In terms of watermark extraction and restoration, the reconstruction quality is clear and without obvious distortion. Therefore, it can be proven that the scheme can simultaneously meet the transmission efficiency and copyright protection requirements of medical images.