Tamper localization and recovery watermarking schemes for medical images in PACS

This thesis focuses on the implementation of medical image watermarking in Picture Archiving and Communications System (PACS). PACS is an important part of information technology infrastructure in a health institution. Medical images stored in PACS are vulnerable to malicious modifications. Watermarking can be used to authenticate medical images and provide the additional security needed on top of the existing security measures that were already in place. Watermarking methods applied to medical images should be reversible or if not, an area known as a region of interest (ROI) needs to be defined on the image to retain the original information. The watermarked image produced should have visual quality similar to its original version. Watermarking schemes should also be tested in a simulated operational environment for practicality verification. In this thesis, three watermarking schemes for medical image are proposed. All three watermarking schemes were tested by watermarking eight different samples. The PSNR of the watermarked images were measured. The watermarked images were tampered by cloning, salt and pepper noise, rotation and smoothing. The tampered images were recovered and the success rates were taken. The first proposed scheme, the reversible tamper localization and recovery scheme (R-TLR). The watermarked image can be reversed by restoring the least significant bits (LSBs) in a predefined region of interest (ROI) and regions of non-interest (RONI) to their original states. The watermarked images have an high average PSNR of 53.9 dB.The success rate of the tamper localization and recovery is close to 100%. Secondly, the tamper localization and lossless recovery (TALLOR) scheme uses compressed recovery information embedded outside of the ROI to achieve exact or lossless recovery after tampering is detected. Thirdly, an enhanced scheme, the tamper localization and lossless recovery with ROI segmentation (TALLOR-RS) is also proposed. It managed to reduce the tamper localization and recovery average processing time by approximately 53%. The average PSNR of the watermarked images for both schemes is high at 48.3 dB and 48.2 dB for TALLOR and TALLOR-RS respectively. The proposed schemes have 100% success rate for tamper localization and recovery. Both schemes also performed better than the scheme developed by Osamah and Khoo(2011). The proposed watermarking schemes were then tested in a simulated PACS environment with good results. A design of a watermark embedder and image authenticator (WEIA) to facilitate the implementation of watermarking in PACS is then proposed. In conclusion, watermarking schemes can be effectively implemented in practice to prevent fraud and improve information security