Recently the active use of Internet and multimedia technologies has increased the violation of copyright. Digital watermarking has a crucial role in the field of multimedia copyright protection. In this paper, an optimized image watermarking technique based on Discrete Wavelet Transform (DWT) and Hadamard transform is proposed. Genetic Algorithm (GA) is an optimizing technique used to satisfy the tradeoff between robustness and imperceptibility. Blind property is conducted using the estimation capability of Decision Tree (DT). First the host image is transformed using second level DWT, then Hadamard transform is applied on selected DWT sub-bands. Adaptive multiple strength values are computed using GA. The watermark is successfully extracted using the trained DT which estimates the original coefficients needed for the extraction process without the need for the original host image. The proposed technique is evaluated against several types of attacks: compression, median filtering, salt and pepper noise, histogram equalization, blurring, scaling, painting, and cropping. Experimental results show that the proposed watermarking technique is robust against these attacks while keeping good imperceptibility. The proposed technique outperforms the compared techniques according to robustness, imperceptibility, and capacity.

最近，Internet和多媒体技术的积极使用增加了对版权的侵犯。数字水印在多媒体版权保护领域具有至关重要的作用。提出了一种基于离散小波变换（DWT）和Hadamard变换的优化图像水印技术。遗传算法（GA）是一种优化技术，用于满足鲁棒性和不可感知性之间的权衡。使用决策树（DT）的估计能力来执行盲属性。首先，使用第二级DWT变换主机图像，然后将Hadamard变换应用于选定的DWT子带。自适应多重强度值使用GA计算。使用训练好的DT成功提取水印，该DT估计提取过程所需的原始系数，而无需原始宿主图像。针对几种攻击类型对提出的技术进行了评估：压缩，中值滤波，盐和胡椒噪声，直方图均衡，模糊，缩放，绘制和裁剪。实验结果表明，所提出的水印技术对这些攻击具有鲁棒性，同时又保持了良好的感知能力。根据鲁棒性，不可感知性和容量，所提出的技术优于已比较的技术。实验结果表明，所提出的水印技术对这些攻击具有鲁棒性，同时又保持了良好的感知能力。根据鲁棒性，不可感知性和容量，所提出的技术优于已比较的技术。实验结果表明，所提出的水印技术对这些攻击具有鲁棒性，同时又保持了良好的感知能力。根据鲁棒性，不可感知性和容量，所提出的技术优于已比较的技术。