用于CMOS图像传感器芯片的多功能盲水印算法

提出了一种在小波域中同时嵌入鲁棒水印和脆弱水印的算法。根据相邻小波系数间的关系在JPEG压缩后大多不发生变化的事实,首先对原始载体图像进行一级小波分解,然后将鲁棒水印嵌入到小波的低频系数LL中,脆弱水印嵌入到小波的高频子带HL中,水印提取和图像认证均不需要原始图像参与,很好实现了水印的盲检测。实验结果表明:本文算法对常见的非恶意攻击提取的鲁棒水印和半脆弱水印的归一化相关系数(NC)值几乎都在0.8以上;对恶意的图像操作可以精确检测和定位;算法嵌入容量大,抗JPEG压缩性能高和计算简单,可以集成于CMOS图像传感器芯片中。

A multifunction blind watermarking algorithm for CMOS image sensor chips

An algorithm is proposed which embeds both robust watermark and semi-fragile watermark into the wavelet domain.Based on the fact that most of the size-relationship between adjacent wavelet coefficients don′t change after JPEG compression,the algorithm performs a wavelet decomposition of the original image firstly;then it embeds the robust watermark into LL sub-band,and the semi-fragile watermark into HL sub-band.Because the original image is not necessary during watermark extraction and image authentication,this algorithm realizes a good implementation of blind watermark detection.Experimental results show that,for common non-malicious attacks,the NC value of extracted robust watermark and semi-fragile watermark with our algorithm is above 0.8 or more;also it can fulfill accurate detection and location for malicious image manipulation.In addition,the algorithm has advantages such as large embedding capacity,good performance against JPEG compression and easy to compute,furthermore the algorithm can be integrated in CMOS image sensor chips.