A Combination of Feature-Points-Based and SVD-Based Image Watermarking Algorithm

A novel image watermarking algorithm against geometric attacks is proposed to solve watermarking synchronization problems. In watermark embedding, firstly, watermark is embedded into host image by a Single Value Decomposition (SVD) method; then, feature-points of watermarked image are extracted using Scale Invariant Feature Transform (SIFT) algorithm. In watermark detection, firstly, the watermarked image which is possibly suffered from geometric attacks is matched with undistorted original watermarked image to find out arbitrary three pairs of matched feature-points; then, using the scale and location relationship of these feature-points, rotation, scaling, and translation that the watermarked image is possibly suffered from could be estimated, and the estimated parameters will be used for watermarking resynchronization. Experiment result shows that the new algorithm performs well in terms of good invisibility and robustness against geometric attacks.