Abstract: This paper presents a new self-embedding watermark scheme for image authentication, which extracts image features from the U component of the SVD domain to generate watermark and embeds watermark in the D component of the SVD domain. In the new scheme, the watermark generating and watermark embedding are disposed in the image itself, and the received image authentication needs no information about the original image or watermark. The scheme increases watermark security and prevents forged watermark. Experimental results show that the algorithm can identify intentional content modification and common image processing, and also indicate the location where a modification takes place.

Key words: authentication, self-embedding watermark, singular value decomposition(SVD), image features

摘要： 提出了一种新颖的用于图像认证的自嵌入水印算法，在图像奇异值分解(SVD)域的U分量中提取图像特征信息生成水印，水印嵌入在SVD域的D分量中，水印的生成和嵌入都在图像本身中进行，图像认证时不需要原始图像和任何有关水印的附加信息，从而提高了水印的安全性和保密性。实验结果表明，该算法能很好地将对图像内容的恶意篡改和常规的图像处理区分开，并给出内容篡改的位置。

关键词: 认证, 自嵌入水印, 奇异值分解, 图像特征

CLC Number: 

• TP391