A novel technique for geometrically robust blind image watermarking extraction

In this paper, we introduce a novel method for extracting information in blind image watermarking based on Scale Invariant Feature Transform (SIFT) and the even-odd quantization algorithm. Unlike some existing methods using the same strategy of selecting the robust non-overlapped regions in both embedding and extraction processes, which may occur some errors in watermark synchronization, we propose a fast effective algorithm exploiting the correlation among all detected regions to recover the embedded information. As a result, our method is not only applied to verify the presence of the embedded data but also allows to extract information directly without the original message. Moreover, in order to obtain the security, a secret key is used associated with the SIFT factors to distribute the embedded regions randomly. The experimental results show that the information extraction of our proposed method is implemented successfully and robustly against geometric attacks of rotation, scaling and cropping.