Abstract: A coordinate system of the original image is established by using the feature points which satisfy with a function relation. After the original image transforms into a standard coordinate system, a rectangle ring watermark is adaptively embedded in Discrete Wavelet Transform (DWT) domain from the center to margin by level and level. Then the feature points are embedded into the image and searched before watermark detection. Regardless of the shape that the host image is cropped, the watermark is detected from center to margin after the watermarked image is transformed into the standard coordinate system. The watermark can be detected without the size of original image. The detection is blind. Experimental results show that the proposed scheme is effectively against general and crop attacks.

Key words: digital watermark, coordinate system, statistical characteristics

摘要： 运用满足函数关系的特征点方法建立源图像的坐标系。源图像移入标准坐标系后，基于小波系数块的统计特性，数字水印从中心到四周分层自适应地嵌入到小波域，并嵌入特征点。水印检测前先检测特征点，重新建立已嵌入水印图像的坐标系作为校准系统。源图像遭受任何剪切，已嵌入水印的图像变换到标准坐标系后，由中心到四周分层检测水印。水印检测无需知道源图像大小。实验结果表明该方法能抵抗一般攻击和任意剪切攻击，包括图像四周全部遭受剪切的攻击。

关键词: 数字水印, 坐标系, 统计特性

CLC Number: 

• TP309