中文摘要：

提出一种小波域中基于小波系数方向树模型的支持向量机自适应水印算法.利用支持向量机在小样本的情况下具有良好的学习和泛化性能的特点,建立小波域内不同尺度上同方向子带内相同空间位置的小波系数间的关系模型;根据关系模型来嵌入和提取水印.为提高图像视觉效果,采用模糊聚类的方法自适应地选取水印的嵌入位置.实验结果表明：该方法对常见的图像攻击具有很好的鲁棒性,而且水印的隐蔽性好、安全性强.

英文摘要：

A DWT domain image watermarking scheme based on the direction-tree structure was proposed,and by using support vector machine（SVM） it became possible to detect the inherent relation between wavelet coefficients in the same direction subband and the same spatial location but different resolution.Due to good learning ability and generalizing ability of SVM under the limited training samples,it can well learn the above relation.According to the relation model,a bit of the watermark was embedded and extracted.To be more imperceptible,the author used fuzzy C-mean clustering algorithm to adaptively determine the watermark embedding locations.Moreover,the watermark recovery did require the original image.Experimental results show that the proposed algorithm had good perceptual quality and high robustness to common image processing operation and the JPEG lossy compression.