一种基于LS-SVM和模板匹配修正的抗几何攻击数字水印技术

A Novel Digital Watermarking Approach against Geometic Attacks Using LS-SVM and Template Matching Correction

数字水印技术是信息隐藏的一个重要方向,传统的数字水印技术对一般攻击的防御能力较强,但几何形变攻击的数字水印在提取时仍然有难度,提出一种基于最小二乘法支持向量机和模板匹配修正的数字水印技术.首先,用分块离散余弦变换进行水印嵌入,然后利用训练后的最小二乘法支持向量机粗修正几何形变参数,针对存在的误差采用模板匹配方法进行精细修正.实验证明的算法不仅可以抵抗传统的攻击比如:噪声、滤波和裁剪,在抵抗几何形变时鲁棒性更好.

It is an important direction of information hiding for digital watermarking. The traditional digital watermarking has stronger defense ability against general attacks,but it is still difficult to geometric attacks. In this paper,the digital watermarking technique based on least squares support vector machines and template matching is proposed. Firstly,DCT is performed on coefficient block and watermark image is embedded. Subsequently,the trained least squares support vector machine is employed to extract the watermark image blindly by using the template matching for correction. Experimental results confirm that the proposed scheme is not only robust against common attacks such as noise,filter and crop,but also robust against geometrical attacks.