Enhancement of robustness and security for feature-based watermarking using wavelets

Abstract

Generally, feature-based image watermarking uses the local image features present in an image. The invariance of the features is exploited for achieving robustness against attacks, but the leakage of information is often unconsidered. Here an image adaptive watermarking approach is proposed which uses invariant feature regions for enhancing the robustness and security. Feature regions are determined using the scale-adapted auto correlation matrix and the Laplacian-ofGaussian operation. The leakage of the information can be controlled with the feature detection procedure. Circular feature regions are determined due to its invariance property. The selected regions are normalized to obtain a geometrically invariant feature region into which the watermark is embedded with the hybrid DWT-SVD algorithm. The evaluation results present the proposed scheme can withstand various attacks which may also include the noise attacks, signal processing attacks and geometric distortion. This results in the enhancement of the robustness.