Enhanced Robustness for Digital Images Using Geometric Attack simulation
J. Ferdin Joe, T. Vaikunda Raja · Procedia Engineering · 2012
The effectiveness of a digital watermarking algorithm is indicated by the robustness of embedded watermarks against various attacks. Attacks which attempt to destroy watermarks have two types, noise-like signal processing and geometric distortions. To select a non-overlapping region, some reference parameters have been used, the regions with higher feature corner response are selected when the Harris–Laplacian or Harris–affine detector is used. The number of neighboring feature points inside a region is a major reference for obtaining a non-overlapping set in the watermarking method based on Mexican Hat wavelet scale interaction. A watermarked feature region may have different degrees of robustness against different techniques if there is prior knowledge of each region's attack resistance capability. We proposed a feature region selection method based on the idea of simulated attacking and finding the Optimal feature region set.