用于CMOS图像传感器芯片的版权保护盲水印算法  

An adaptive blind watermarking algorithm for copyright protection in CMOS image sensor chips

作　　者：于平平[1] 姚素英[1] 于俊庭[1] 徐江涛[1] 常晔 厉果育 

机构地区：[1]天津大学电子信息工程学院,天津300072 [2]山东省枣庄市政府信息中心,277101

出　　处：《光电子．激光》2010年第4期579-583,共5页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金资助项目(60806010);教育部博士点新教师基金资助项目(200800561111)

摘　　要：提出了一种易于硬件实现的、用于图像版权保护的自适应盲水印算法。利用R、G和B分量在图像信息中各自的不同特点做不同处理;根据人眼的视觉特性,通过对图像G分量纹理强度分析自适应地选择合适的水印拉伸因子,从R分量中提取图像内容来加密水印,对B分量进行DCT变换来嵌入不可见盲水印。采用伪随机序列发生器实现水印置乱,基于DCT中低频系数求均值来嵌入盲水印,大大降低了硬件消耗。实验结果表明,算法对噪声干扰、滤波、锐化、剪裁和模糊处理具有很好的鲁棒性,提取出的水印的归一化相关系数均在0.5之上,大于人眼可识别的检测阈值。An adaptive blind watermarking algorithm is presented for copyright protection of color image, which can be readily implemented in hardware. According to the different characteristics of R,G and B components included in image information, this algorithm performs different process. Based on the property of human visual system（HVS）, the watermark stretch factor can be adaptively chosen through ana- lyzing the texture intensity of G-component. Then the watermark is encrypted to the image content extracted from R-component. And after the DCT transformation,imperceptible blind watermarking is em- bedded into the B-component of the image information. In order to achieve the watermark scrambling, the pseudo-random sequence generator is used. The hardware comsumption is significantly reduced by em- bedding blind watermarking based on the mean of the low-frequency coefficients in DCT domain. Experimental results show that the proposed algorithm is effective and robust for noise jamming, filtering, sharpening, cropping and fuzzy processing. The normalized correlation coefficients extracted from the wa- termarked images are all above 0. 5,which are greater than the detection threshold that human eye can distinguish.

关 键 词：盲水印 版权保护 DCT 硬件实现 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]