An image watermarking technique with improved resistance to geometric distortions using image feature points

摘要

A blind image watermarking technique with improved robustness to geometric distortions is proposed. The technique relies on geometric properties of image feature points, which are extracted using a feature point extractor based on scale-interaction of Mexican-hat wavelets. These feature points are used to obtain geometric normalization parameters during the embedding process which are used to compensate for geometric distortions during the watermark detection process. Experimental results illustrate the robustness of the proposed technique to geometric attacks in addition to several common image-processing operations.