Simulated attack based feature region selection for efficient digital image watermarking

An ideal feature region selection method for efficient digital image watermarking is proposed in this paper. The aim of this method is to select a feature region set with high efficiency. The feature region sets thus obtained are non-overlapping and has the greatest strength against various attacks. Image quality can be conserved as much as possible after watermarking. Predefined attacks are used to calculate the strength of every candidate feature region. Simulated attacking procedures are used for this purpose. According to the evaluation results, it then adopts procedure to search. First of all we search for a primary feature set which can thwart the most predefined attacks. This is done using track-with-pruning procedure. Some auxiliary feature regions are added to the primary feature region. This is done in order to enhance its resistance to undefined attacks. While doing these procedures the quality of the image shouldn't decrease. This work is modelled as a multidimensional knapsack problem and is solved using genetic algorithm. In the Experiment Stir Mark attacks are applied on some test images. The result supports our expectation that the primary feature set can thwart all the predefined attacks. The extension of the primary feature set can enhance the strength against undefined attacks. The proposed method shows better performance in efficient digital watermarking compared with some well-known feature based method.