Robust Blind Watermarking against Geometrical Attacks with Medical Images for EMR/EHR

Abstract

In this paper, an efficient solution for robust watermarking against geometrical attacks with medical images is proposed. In the embedding process, at first, we select the robust non-overlapping Scale-Invariant Feature Transform regions. Then, we embed the same information into these regions by using the even-odd quantization technique. In the extraction process, blind recovery of information is performed by exploiting the correlation among all detection regions. Both fan-shaped and half-ring-shaped regions are suggested for multi-bit watermarking. The embedded image quality is analyzed and compared to other image watermarking approaches based on Spread Spectrum or Least Significant Bit techniques. The reliability and the robustness are investigated with various geometrical attacks such as rotation, scaling and cropping. Moreover, our solution is also better than existing methods in terms of usefulness and security for electronic medical/health record applications.