Image watermarking through attack simulation based on block classification

A block-based image classification-based technique is developed in this article to improve our previous simulated attack-based watermarking system. By taking advantage of characteristics of the human visual system, the domain coefficients in block image classification are determined by obtaining the maximum hidden information of embedded watermarks so that the difference in image quality between block-based watermarks and simulated attack watermarks is hardly to be detected and a better robustness of the embedded watermarks can be achieved. The experimental results demonstrate that the proposed system can resist various common attacks such as JPEG compression, smoothing, sharpening, cropping, and noise at the expense of a small amount of loss in the quality of embedded watermarks due to simulated attacks.