Robust Blind Medical Image Watermarking Using Quantization and SIFT with Enhanced Security

The paper proposes an efficient blind robust watermarking solution for medical images based on a combination of the Scale Invariant Feature Transform (SIFT) and even-odd quantization. Unlike most existing methods using SIFT with original image, our proposed algorithm can extract the embedded information without original image by selecting only non-overlapping features in embedding process and exploiting the correlation among all detecting regions. As a result, both detection and extraction of embedded information can be obtained with our method. Moreover, it can be expanded to multi-bit watermarking with two suggestions of fan-shaped and half-ring-shaped regions. The experimental results are implemented with various medical images and evaluated about the quality, the reliability and the robustness against common medical image processing attacks including filtering, compression, rotation, scaling and cropping. Furthermore, the security in embedding and extracting information is also enhanced in our solution.