A novel chaotic IWT-LSB blind watermarking approach with flexible capacity for secure transmission of authenticated medical images

Abstract

Nowadays, secure medical data transmission is an essential issue in telemedicine through unsecured channels. Watermarking methods are widely used to provide optimal security for medical image transmission. In this method, a secure and blind watermarking algorithm based on Integer Wavelet Transform (IWT), Least Significant Bit (LSB), and chaotic sequences (ILC) with a high capacity for grayscale medical and non-medical images is ILC. In this scheme, the high-level security is achieved by chaotic sequences that are applied for watermark encryption and determining the location of host blocks and host coefficients in the embedding process. At first, the cover medical image is divided into two distinct parts: the region of interest (ROI) and the region of non-interest (RONI) areas. Due to the high sensitivity of the ROI data in disease diagnosis, no embedding is done in this area. An ROI integrity check data (ICR), produced by some significant features from ROI, to detect tampered blocks inside this area. The main watermark, which is the encryption of ICR along with the patient’s personal and medical information, is embedded in the middle-frequency at the first and second levels of IWT sub-bands. Furthermore, to verify the watermark, integrity check data (ICW) is produced and sent along with other essential data to the receiver. For identifying authentication purposes, the physician’s signature is embedded at the third level (LL3) sub-band of IWT which has the highest energy and robustness. Moreover, the algorithm is flexible for using different sizes of text watermarks and it guarantees a high payload capacity. It is noteworthy the maximum lengths of text watermarks used in the method for 512×512 and 1024×1024 images are 47,784 and 164,620 bits, respectively. Although the size of the embedded watermark increases with the size of the ROI area, the image transparency remains almost constant and also the watermark is extracted with a bit error rate of zero. By doing various analyses, the experimental results represent superior imperceptibility by an average PSNR value of 75 dB. This method provides high imperceptibility and robustness against various attacks compared with state-of-the-art algorithms.