A Robust Watermarking Algorithm Based on Salient Image Features

A feature-based robust watermarking algorithm against geometric transformation is proposed in this paper. As salient features in the image are relatively stable under geometric transformation, they may serve as reference points to synchronize the processes of watermark embedding and detection. To further increase the repeatability of feature detection, a novel feature enhancement technique is proposed. Such enhancement is crucial because most existent watermarking algorithms suffer from lost of synchronization due to image attacks, and the watermark would not be detected correctly if the repeatability rate is low. Image moments were also used to achieve geometric invariance of watermark embedding and detection. Experimental results demonstrated that our algorithm can survive various geometric transformation and common image processing operations.