一种鲁棒性好的图像认证新方法

A Rubost Image Authentication Algorithm

本文提出了一种新的图像认证方法，这种方法能够区分图像是被恶意篡改还是无意篡改。图像认证是检测图像是否被恶意篡改，认证图像数据的完整性，同时能容忍压缩或噪声对原图像质量造成影响的一种新方法。本文通过提取图像的直方圆不变矩，作为认证码，然后利用奇异值分解的方法将这些认证码作为水印嵌入到原图像。实验证明，追种方法不仅适用于BMP图像篡改检测，还能比较准确地检测出被JPEG压缩后的图像是否被篡改，并可准确确定图像被篡改的位置，同时该方法抗几何失真性强。

The goal of this paper is to design a watermarking scheme capable of distinguishing visible but nonmalicious changes due to common image processing operation from malicious changes. Image authentication is to verify of an image by discovering malicious manipulation. Small modification such as lossy compression aud noises does not affect the results of image authentication. This paper presents an image authentication ,algorithm which utilizes moment invariants of image histogram. First, moment invariants of the image histogram are extracted, then, by using singular value decomposition （SVD） , those moment invariants which are equivalent to digital watermarking are embedded into the original image. Experimental results show that this algorithm can both detect BMP image and JPEG compressed image whether it is tempered and ascertain where the image is tempered exactly.